Skip to contentSkip to conversation

Cleanup Octeon DWC3 glue code

PATCH Superseded Ladislav Michl 4 participants 27 messages Jun 19, 2023 – Jul 5, 2023

Subsystems (3)

v1 of 5 Superseded Latest is v5 →

lore
Share & tools
lore
#0 lore raw

cover Cleanup Octeon DWC3 glue code

Hi,

primary motivation was to fix issue Id 29206 as described in OCTEON III CN70XX/CN71XX Known Issues, version: 1.9. Said document is marked as Marvell Proprietary and Confidential, therefore I'm not sure if I can cite from it.

This probably does not matter too much as the root of the information listed there is a workaround being implemented in OCTEON SDK 3.1.2 patch 7 and later in u-boot/drivers/usb/host/xhci-octeon.c:dwc3_uphy_pll_reset()

My coleague ported that patch to linux-4.9 and it is appended to this cover letter for a reference.

The glue code currently lives in arch/mips/cavium-octeon/octeon-usb.c and loops for each "cavium,octeon-7130-usb-uctl" compatible. However there is no bond with dwc3 core code, so if anything goes wrong in glue code, the loop breaks, leaving dwc3 in reset.

Later on when dwc3 core tries to read any device register, bus error is emited, leading to kernel panic.

Therefore glue code is cleaned-up first, then moved to drivers/usb/dwc3 and modified to match current state of art.

Now it is easier to properly implement said errata. Does it belong to core code as a quirk? Comments and suggestions welcome.

Ladislav Michl (11):
MIPS: OCTEON: octeon-usb: add all register offsets
MIPS: OCTEON: octeon-usb: use bitfields for control register
MIPS: OCTEON: octeon-usb: use bitfields for host config register
MIPS: OCTEON: octeon-usb: use bitfields for shim register
MIPS: OCTEON: octeon-usb: move gpio config to separate function
MIPS: OCTEON: octeon-usb: introduce dwc3_octeon_{read,write}q
MIPS: OCTEON: octeon-usb: cleanup divider calculation
usb: dwc3: Move Octeon glue code from arch/mips
usb: dwc3: dwc3-octeon: Convert to glue driver
usb: dwc3: dwc3-octeon: Move node parsing into driver probe
usb: dwc3: Add SPDX header and copyright

 arch/mips/cavium-octeon/Makefile          |   1 -
 arch/mips/cavium-octeon/octeon-platform.c |   1 -
 arch/mips/cavium-octeon/octeon-usb.c      | 548 ----------------------
 drivers/usb/dwc3/Kconfig                  |   9 +
 drivers/usb/dwc3/Makefile                 |   1 +
 drivers/usb/dwc3/dwc3-octeon.c            | 545 +++++++++++++++++++++
 drivers/usb/dwc3/dwc3-of-simple.c         |   1 -
 7 files changed, 555 insertions(+), 551 deletions(-)
 delete mode 100644 arch/mips/cavium-octeon/octeon-usb.c
 create mode 100644 drivers/usb/dwc3/dwc3-octeon.c
---
 arch/mips/cavium-octeon/octeon-usb.c | 592 +++++++++++++++++++++++++++
 1 file changed, 592 insertions(+)
arch/mips/cavium-octeon/octeon-usb.c +592 -0
index 514af27e23ff..544c2db6c6c3 100644
@@ -14,7 +14,9 @@
 #include <linux/delay.h>
 #include <linux/of_platform.h>
 #include <linux/io.h>
+#include <linux/jiffies.h>
 
+#include <asm/octeon/cvmx-usbdrdx-defs.h>
 #include <asm/octeon/octeon.h>
 
 /* USB Control Register */
@@ -225,6 +227,371 @@ static DEFINE_MUTEX(dwc3_octeon_clocks_mutex);
 static uint8_t clk_div[OCTEON_H_CLKDIV_SEL] = {1, 2, 4, 6, 8, 16, 24, 32};
 
 
+/**
+ * Reset PLL source on driver probe.
+ * Patch taken from: u-boot/drivers/usb/host/xhci-octeon.c:octeon3_usb_clocks_start
+ */
+
+/** Internal indirect UPHY register that controls the power to the UPHY PLL. */
+#define DWC3_INT_IND_UPHY_PLL_PU			0x2012
+/** Internal indirect UPHY PLL register */
+#define DWC3_INT_IND_UPHY_PLL_RESET			0x201C
+/**
+ * Internal indirect register that reports if the phy PLL has lock.  This will
+ * be 1 if lock, 0 if no lock.
+ */
+#define DWC3_INT_IND_PLL_LOCK_REG			0x200b
+
+/** Write enable bit */
+#define DWC3_INT_IND_UPHY_PLL_RESET_WE			(1 << 4)
+/** VCO reset bit */
+#define DWC3_INT_IND_UPHY_PLL_RESET_VCO_RST		(1 << 0)
+/** Write enable bit for DWC3_INT_IND_PLL_POWER_CTL */
+#define DWC3_INT_IND_UPHY_PLL_PU_WE			0x20
+/** Power enable bit for DWC3_INT_IND_PLL_POWER_CTL */
+#define DWC3_INT_IND_UPHY_PLL_PU_POWER_EN		0x02
+
+
+/**
+ * Indirectly read internal DWC3 UPHY registers
+ *
+ * @param	node	CPU node number
+ * @param	index	Port index
+ * @param	addr	Address of register to read
+ *
+ * @return	16-bit register value
+ */
+uint32_t dwc3_uphy_indirect_read(struct device *dev, int node, int index, uint32_t addr)
+{
+	union cvmx_usbdrdx_uctl_portx_cr_dbg_cfg dbg_cfg;
+	union cvmx_usbdrdx_uctl_portx_cr_dbg_status status;
+	unsigned long start;
+	uint16_t data = 0xffff;
+
+
+	/* 1 */
+	dbg_cfg.u64 = 0;
+	dbg_cfg.s.data_in = addr;
+	cvmx_write_csr_node(node, CVMX_USBDRDX_UCTL_PORTX_CR_DBG_CFG(0, index),
+			    dbg_cfg.u64);
+	/* 2 */
+	dbg_cfg.s.cap_addr = 1;
+	cvmx_write_csr_node(node, CVMX_USBDRDX_UCTL_PORTX_CR_DBG_CFG(0, index),
+			    dbg_cfg.u64);
+	/* 3 */
+	start = jiffies;
+	do {
+		status.u64 = cvmx_read_csr_node(node,
+				CVMX_USBDRDX_UCTL_PORTX_CR_DBG_STATUS(0, index));
+	} while (status.s.ack != 1 && jiffies< (start + HZ));
+	if (status.s.ack != 1) {
+		dev_warn(dev, "%s(%d, %d, 0x%x): status ack = 1 timeout\n", __func__,
+			 node, index, addr);
+		return data;
+	}
+
+	/* 4 */
+	dbg_cfg.u64 = 0;
+	cvmx_write_csr_node(node, CVMX_USBDRDX_UCTL_PORTX_CR_DBG_CFG(0, index),
+			    dbg_cfg.u64);
+	/* 5 */
+	start = jiffies;
+	do {
+		status.u64 = cvmx_read_csr_node(node,
+				CVMX_USBDRDX_UCTL_PORTX_CR_DBG_STATUS(0, index));
+	} while (status.s.ack != 0 && jiffies< (start + HZ));
+	if (status.s.ack) {
+		dev_warn(dev, "%s(%d, %d, 0x%x): status ack = 0 timeout\n", __func__,
+			 node, index, addr);
+		return data;
+	}
+
+	/* 6 */
+	dbg_cfg.s.read = 1;
+	cvmx_write_csr_node(node, CVMX_USBDRDX_UCTL_PORTX_CR_DBG_CFG(0, index),
+			    dbg_cfg.u64);
+	/* 7 */
+	start = jiffies;
+	do {
+		status.u64 = cvmx_read_csr_node(node,
+				CVMX_USBDRDX_UCTL_PORTX_CR_DBG_STATUS(0, index));
+	} while (status.s.ack != 1 && jiffies< (start + HZ));
+	if (status.s.ack != 1) {
+		dev_warn(dev, "%s(%d, %d, 0x%x): Status ack = 0 get data timeout\n",
+			 __func__, node, index, addr);
+		return 0xffff;
+	}
+	/* 8 */
+	data = status.s.data_out;
+	/* 9 */
+	dbg_cfg.u64 = 0;
+	cvmx_write_csr_node(node, CVMX_USBDRDX_UCTL_PORTX_CR_DBG_CFG(0, index),
+			    dbg_cfg.u64);
+	/* 10 */
+	start = jiffies;
+	do {
+		status.u64 = cvmx_read_csr_node(node,
+				CVMX_USBDRDX_UCTL_PORTX_CR_DBG_STATUS(0, index));
+	} while (status.s.ack != 0 && jiffies< (start + HZ));
+	if (status.s.ack != 0) {
+		dev_warn(dev, "%s(%d, %d, 0x%x): Final ack timeout\n", __func__,
+			 node, index, addr);
+		return 0xffff;
+	}
+
+	return data;
+}
+
+/**
+ * Indirectly write internal DWC3 UPHY registers
+ *
+ * @param	node	CPU node number
+ * @param	index	Port index
+ * @param	addr	Address of register to read
+ *
+ * @return	0 for success
+ */
+static uint32_t dwc3_uphy_indirect_write(struct device *dev, int node, int index,
+					   uint32_t addr, uint16_t value)
+{
+	union cvmx_usbdrdx_uctl_portx_cr_dbg_cfg dbg_cfg;
+	union cvmx_usbdrdx_uctl_portx_cr_dbg_status status;
+	unsigned long start;
+
+	/* 1 */
+	dbg_cfg.u64 = 0;
+	dbg_cfg.s.data_in = addr;
+	cvmx_write_csr_node(node, CVMX_USBDRDX_UCTL_PORTX_CR_DBG_CFG(0, index),
+			    dbg_cfg.u64);
+	/* 2 */
+	dbg_cfg.s.cap_addr = 1;
+	cvmx_write_csr_node(node, CVMX_USBDRDX_UCTL_PORTX_CR_DBG_CFG(0, index),
+			    dbg_cfg.u64);
+	/* 3 */
+	start = jiffies;
+	do {
+		status.u64 = cvmx_read_csr_node(node,
+						CVMX_USBDRDX_UCTL_PORTX_CR_DBG_STATUS(0, index));
+	} while (status.s.ack != 1 && jiffies< (start + HZ));
+	if (status.s.ack != 1) {
+		dev_warn(dev, "%s(%d, %d, 0x%x, 0x%x): setting address timed out\n",
+			 __func__, node, index, addr, value);
+		return -1;
+	}
+	/* 4 */
+	dbg_cfg.u64 = 0;
+	cvmx_write_csr_node(node, CVMX_USBDRDX_UCTL_PORTX_CR_DBG_CFG(0, index),
+			    dbg_cfg.u64);
+	/* 5 */
+	start = jiffies;
+	do {
+		status.u64 = cvmx_read_csr_node(node,
+						CVMX_USBDRDX_UCTL_PORTX_CR_DBG_STATUS(0, index));
+	} while (status.s.ack != 0 && jiffies< (start + HZ));
+	if (status.s.ack != 0) {
+		dev_dbg(dev, "%s(%d, %d, 0x%x, 0x%x): Timeout waiting for ack = 0\n",
+			__func__, node, index, addr, value);
+		return -1;
+	}
+
+	/* 6 */
+	dbg_cfg.u64 = 0;
+	dbg_cfg.s.data_in = value;
+	cvmx_write_csr_node(node, CVMX_USBDRDX_UCTL_PORTX_CR_DBG_CFG(0, index),
+			    dbg_cfg.u64);
+	/* 7 */
+	dbg_cfg.s.cap_data = 1;
+	cvmx_write_csr_node(node, CVMX_USBDRDX_UCTL_PORTX_CR_DBG_CFG(0, index),
+			    dbg_cfg.u64);
+	/* 8 */
+	start = jiffies;
+	do {
+		status.u64 = cvmx_read_csr_node(node,
+				CVMX_USBDRDX_UCTL_PORTX_CR_DBG_STATUS(0, index));
+	} while (status.s.ack != 1 && jiffies< (start + HZ));
+	if (status.s.ack != 1) {
+		dev_dbg(dev, "%s(%d, %d, 0x%x, 0x%x): Timeout setting data, waiting for ack = 1\n",
+			__func__, node, index, addr, value);
+		return -1;
+	}
+	/* 9 */
+	dbg_cfg.u64 = 0;
+	cvmx_write_csr_node(node, CVMX_USBDRDX_UCTL_PORTX_CR_DBG_CFG(0, index),
+			    dbg_cfg.u64);
+	/* 10 */
+	start = jiffies;
+	do {
+		status.u64 = cvmx_read_csr_node(node,
+				CVMX_USBDRDX_UCTL_PORTX_CR_DBG_STATUS(0, index));
+	} while (status.s.ack != 0 && jiffies< (start + HZ));
+	if (status.s.ack != 0) {
+		dev_dbg(dev, "%s(%d, %d, 0x%x, 0x%x): Timeout waiting for ack = 0\n",
+			__func__, node, index, addr, value);
+		return -1;
+	}
+	/* 11 */
+	dbg_cfg.u64 = 0;
+	dbg_cfg.s.write = 1;
+	cvmx_write_csr_node(node, CVMX_USBDRDX_UCTL_PORTX_CR_DBG_CFG(0, index),
+			    dbg_cfg.u64);
+	/* 12 */
+	start = jiffies;
+	do {
+		status.u64 = cvmx_read_csr_node(node,
+				CVMX_USBDRDX_UCTL_PORTX_CR_DBG_STATUS(0, index));
+	} while (status.s.ack != 1 && jiffies< (start + HZ));
+	if (status.s.ack != 1) {
+		dev_warn(dev, "%s(%d, %d, 0x%x, 0x%x): Timeout waiting for write ack = 1\n",
+			 __func__, node, index, addr, value);
+		return -1;
+	}
+	/* 13 */
+	dbg_cfg.u64 = 0;
+	cvmx_write_csr_node(node, CVMX_USBDRDX_UCTL_PORTX_CR_DBG_CFG(0, index),
+			    dbg_cfg.u64);
+	/* 14 */
+	start = jiffies;
+	do {
+		status.u64 = cvmx_read_csr_node(node,
+				CVMX_USBDRDX_UCTL_PORTX_CR_DBG_STATUS(0, index));
+	} while (status.s.ack != 0 && jiffies< (start + HZ));
+	if (status.s.ack != 0) {
+		dev_warn(dev, "%s(%d, %d, 0x%x, 0x%x): Timeout waiting for write ack = 0\n",
+			 __func__, node, index, addr, value);
+		return -1;
+	}
+
+	return 0;
+}
+
+/**
+ * Performs a full reset of the UPHY PLL.  Note that this is normally done
+ * internally by a state machine when the UPHY is brought out of reset but this
+ * version gives far more time for things to settle before continuing.
+ *
+ * @param node	CPU node
+ * @param index	XHCI controller index
+ *
+ * @return 0 for success, 1 if UPHY PLL does not have lock.
+ */
+static int dwc3_uphy_pll_reset(struct device *dev, int node, int index)
+{
+	const uint16_t pwr_reg = DWC3_INT_IND_UPHY_PLL_PU;
+	const uint16_t ctrl_reg = DWC3_INT_IND_UPHY_PLL_RESET;
+	const uint16_t lock_reg = DWC3_INT_IND_PLL_LOCK_REG;
+	uint16_t ctrl_val, pwr_val;
+
+	dev_dbg(dev, "%s(%d, %d)\n", __func__, node, index);
+
+	/* 1. Turn on write enable so we can assert reset to the PLL VCO */
+	ctrl_val = dwc3_uphy_indirect_read(dev, node, index, ctrl_reg);
+	ctrl_val |= DWC3_INT_IND_UPHY_PLL_RESET_WE;
+	dwc3_uphy_indirect_write(dev, node, index, ctrl_reg, ctrl_val);
+	mdelay(1);
+
+	/* 2. Turn on write enable for PLL power control */
+	pwr_val = dwc3_uphy_indirect_read(dev, node, index, pwr_reg);
+	pwr_val |= DWC3_INT_IND_UPHY_PLL_PU_WE;
+	dwc3_uphy_indirect_write(dev, node, index, pwr_reg, pwr_val);
+	mdelay(1);
+
+	/* 3. Assert VCO reset */
+	ctrl_val |= DWC3_INT_IND_UPHY_PLL_RESET_VCO_RST;
+	dwc3_uphy_indirect_write(dev, node, index, ctrl_reg, ctrl_val);
+
+	mdelay(1);
+
+	/* 4. Power down the PLL */
+	pwr_val &= ~DWC3_INT_IND_UPHY_PLL_PU_POWER_EN;
+	dwc3_uphy_indirect_write(dev, node, index, pwr_reg, pwr_val);
+	mdelay(1);
+
+	/* 5. Power on the PLL while VCO is held in reset */
+	pwr_val |= DWC3_INT_IND_UPHY_PLL_PU_POWER_EN;
+	dwc3_uphy_indirect_write(dev, node, index, pwr_reg, pwr_val);
+
+	/* Wait for things to stabilize before taking VCO out of reset */
+	mdelay(1);
+
+	/* 6. Take the VCO out of reset */
+	ctrl_val &= ~DWC3_INT_IND_UPHY_PLL_RESET_VCO_RST;
+	dwc3_uphy_indirect_write(dev, node, index, ctrl_reg, ctrl_val);
+	mdelay(1);
+
+	/* 7. Put the VCO back in reset */
+	ctrl_val |= ~DWC3_INT_IND_UPHY_PLL_RESET_VCO_RST;
+	dwc3_uphy_indirect_write(dev, node, index, ctrl_reg, ctrl_val);
+	mdelay(1);
+
+	/* 8. Power down the PLL */
+	pwr_val &= ~DWC3_INT_IND_UPHY_PLL_PU_POWER_EN;
+	dwc3_uphy_indirect_write(dev, node, index, pwr_reg, pwr_val);
+	mdelay(1);
+
+	/* 9. Power on the PLL while VCO is held in reset */
+	pwr_val |= DWC3_INT_IND_UPHY_PLL_PU_POWER_EN;
+	dwc3_uphy_indirect_write(dev, node, index, pwr_reg, pwr_val);
+	mdelay(1);
+
+	/* 10. Take the VCO out of reset */
+	ctrl_val &= ~DWC3_INT_IND_UPHY_PLL_RESET_VCO_RST;
+	dwc3_uphy_indirect_write(dev, node, index, ctrl_reg, ctrl_val);
+	mdelay(1);
+
+	/* 11. Turn off write enables */
+	pwr_val &= ~DWC3_INT_IND_UPHY_PLL_PU_WE;
+	dwc3_uphy_indirect_write(dev, node, index, pwr_reg, pwr_val);
+
+	ctrl_val &= DWC3_INT_IND_UPHY_PLL_RESET_WE;
+	dwc3_uphy_indirect_write(dev, node, index, ctrl_reg, ctrl_val);
+
+	mdelay(1);
+
+	/* Return if we have lock or not */
+	return !(dwc3_uphy_indirect_read(dev, node, index, lock_reg) & 1);
+}
+
+/**
+ * Converts a CPU NODE and CSR address to a new address
+ *
+ * @param node	CPU node number
+ * @param addr	CSR address
+ *
+ * @return New CSR address for the specified CPU node
+ */
+static uint64_t node_to_addr(int node, uint64_t addr)
+{
+	return (addr & ~CVMX_NODE_IO_MASK) |
+		((uint64_t)node & CVMX_NODE_MASK) << CVMX_NODE_IO_SHIFT;
+}
+
+/**
+ * Writes to a USBDRDX_UAHC register, taking care of any endian conversions.
+ *
+ * @param addr	Address of register to write to
+ * @param val	Value to write to register
+ */
+static void write_uahc_reg(int node, uint64_t addr, uint32_t val)
+{
+	uint64_t addr64 = node_to_addr(node, addr);
+	cvmx_write64_uint32(addr64, cpu_to_le32(val));
+}
+
+/**
+ * Reads from a USBDRDX_UAHC register.  All UAHC registers are in little-endian
+ * format.
+ *
+ * @param addr	Address of register to read
+ * @return value of register
+ */
+static uint32_t read_uahc_reg(int node, uint64_t addr)
+{
+	uint64_t addr64 = node_to_addr(node, addr);
+	return le32_to_cpu(cvmx_read64_uint32(addr64));
+}
+
+#endif
+
 static int dwc3_octeon_config_power(struct device *dev, u64 base)
 {
 #define UCTL_HOST_CFG	0xe0
@@ -481,6 +848,7 @@ static void __init dwc3_octeon_set_endian_mode(u64 base)
 #define CVMX_USBDRDX_UCTL_CTL(index)				\
 		(CVMX_ADD_IO_SEG(0x0001180068000000ull) +	\
 		((index & 1) * 0x1000000ull))
+
 static void __init dwc3_octeon_phy_reset(u64 base)
 {
 	union cvm_usbdrd_uctl_ctl uctl_ctl;
@@ -491,6 +859,224 @@ static void __init dwc3_octeon_phy_reset(u64 base)
 	cvmx_write_csr(CVMX_USBDRDX_UCTL_CTL(index), uctl_ctl.u64);
 }
 
+static int __init dwc3_octeon_device_pll_lock(struct device *dev, u64 base)
+{
+	union cvmx_usbdrdx_uctl_ctl uctl_ctl;
+	union cvmx_usbdrdx_uahc_gusb2phycfgx phycfg;
+	union cvmx_usbdrdx_uahc_gusb3pipectlx pipectl;
+	union cvmx_usbdrdx_uahc_gctl gctl;
+	union cvmx_usbdrdx_uahc_gsts gsts;
+	union cvmx_usbdrdx_uahc_portscx portsc;
+
+	uint64_t phycfg_addr;
+	uint64_t pipectl_addr;
+	uint64_t gctl_addr;
+	uint64_t gsts_addr;
+
+	int index, ret;
+	int node = 0;
+
+	switch(base) {
+	case 0x8001180068000000ull:
+		index = 0;
+		break;
+	case 0x8001180069000000ull:
+		index = 1;
+		break;
+	default:
+		dev_err(dev, "Invalid base address 0x%llx for node %d.", base, node);
+		return -1;
+	}
+
+	/* From previous steps */
+	/* Step 13, wait 10uS for the PHY to complete its reset */
+	udelay(10);
+
+	/* 14. Deassert UPHY reset */
+	uctl_ctl.u64 = cvmx_read_csr_node(node,
+					  CVMX_USBDRDX_UCTL_CTL(index));
+	uctl_ctl.s.uphy_rst = 0;
+	cvmx_write_csr_node(node, CVMX_USBDRDX_UCTL_CTL(index),
+			    uctl_ctl.u64);
+	/* Setp 15, wait for at least 45us for UPHY to output stable
+	 * PHYCLOCK
+	 */
+	udelay(50);
+
+	/* Step 16: Initialize any other strap signals necessary and
+	 * make sure they propagate by reading back the last register
+	 * written.
+	 *
+	 * a. UCTL:
+	 *	uctl_port0_cfg_*[*_tune]
+	 *	uctl_port0_cfg_*[pcs_*]
+	 *	uctl_port0_cfg_*[lane0_tx_term_offset]
+	 *	uctl_port0_cfg_*[tx_vboost_lvl]
+	 *	uctl_port0_cfg_*[los_bias]
+	 *	uctl_host_cfg
+	 *	uctl_shim_cfg
+	 */
+	/* b. UAHC: only the following UAHC registers are accessible
+	 *	    during CoreSoftReset.
+	 * - USBDRDX_UAHC_GCTL
+	 * - USBDRDX_UAHC_GUCTL
+	 * - USBDRDX_UAHC_GSTS
+	 * - USBDRDX_UAHC_GUID
+	 * - USBDRDX_UAHC_GUSB2PHYCFG
+	 * - USBDRDX_UAHC_GUSB3PIPECTL
+	 */
+
+	/* 17. Release soft reset the UPHY and UAHC logic via the UAHC
+	 *     controls.
+	 */
+
+	phycfg_addr = CVMX_USBDRDX_UAHC_GUSB2PHYCFGX(0, index);
+	pipectl_addr = node_to_addr(node,
+				CVMX_USBDRDX_UAHC_GUSB3PIPECTLX(0, index));
+	gctl_addr = CVMX_USBDRDX_UAHC_GCTL(index);
+	gsts_addr = CVMX_USBDRDX_UAHC_GSTS(index);
+
+	phycfg.u32 = read_uahc_reg(node, phycfg_addr);
+	phycfg.s.physoftrst = 1;
+	/* Disable low power mode due to mcbuggin 29646 */
+	phycfg.s.enblslpm = 0;
+	write_uahc_reg(node, phycfg_addr, phycfg.u32);
+
+	pipectl.u32 = read_uahc_reg(node, pipectl_addr);
+	pipectl.s.physoftrst = 1;
+	/* Disable low power mode due to mcbuggin 29646 */
+	pipectl.s.suspend_en = 0;
+	write_uahc_reg(node, pipectl_addr, pipectl.u32);
+
+	udelay(100);
+
+	gctl.u32 = read_uahc_reg(node, gctl_addr);
+	gctl.s.coresoftreset = 1;
+	write_uahc_reg(node, gctl_addr, gctl.u32);
+
+#ifdef DEBUG
+	if (!dwc3_uphy_indirect_read(dev, node, index,
+				     DWC3_INT_IND_PLL_LOCK_REG))
+		dev_warn(dev, "Detected no lock on USB 2 PHY PLL\n");
+#endif
+
+	dev_warn(dev, "DWC3: usb pll locked %u.",
+		 dwc3_uphy_indirect_read(dev, node, index, DWC3_INT_IND_PLL_LOCK_REG));
+	/* 19. Reset the USB PHY PLL.  This phy is reset during the
+	 *     UPHY reset process but the internal state machine seems
+	 *     to be too fast and sometimes the PLL Does not come up.
+	 *     The software reset sequence has much longer delays
+	 *     which allows things to settle properly before proceeding.
+	 */
+
+	dev_warn(dev, "%s: Resetting USB PHY PLL\n", __func__);
+	ret = dwc3_uphy_pll_reset(dev, node, index);
+	if (ret) {
+		dev_warn(dev, "%s: UPHY PLL on node: %d, interface: %d not locked after reset.\n",
+			 __func__, node, index);
+	}
+	mdelay(1);
+
+	/* 18. Configure the remaining UAHC_G* registers including any
+	 *     that were not configured in step 16.-b
+	 */
+#ifdef DEBUG
+	if (!dwc3_uphy_indirect_read(dev, node, index,
+				     DWC3_INT_IND_PLL_LOCK_REG))
+		dev_warn(dev, "Detected no lock on USB 2 PHY PLL\n");
+#endif
+
+	/* 19. Reset the USB PHY PLL.  This phy is reset during the
+	 *     UPHY reset process but the internal state machine seems
+	 *     to be too fast and sometimes the PLL Does not come up.
+	 *     The software reset sequence has much longer delays
+	 *     which allows things to settle properly before proceeding.
+	 */
+
+	dev_info(dev, "%s: Resetting USB PHY PLL\n", __func__);
+	ret = dwc3_uphy_pll_reset(dev, node, index);
+	if (ret) {
+		dev_warn(dev, "%s: UPHY PLL on node: %d, interface: %d not locked after reset.\n",
+			 __func__, node, index);
+		return -1;
+	}
+
+	dev_warn(dev, "DWC3: usb pll locked %u after reset.",
+		 dwc3_uphy_indirect_read(dev, node, index, DWC3_INT_IND_PLL_LOCK_REG));
+
+	pipectl.u32 = read_uahc_reg(node, pipectl_addr);
+	dev_info(dev, "%s: GUSB3PIPECTL: 0x%x\n", __func__, pipectl.u32);
+
+	pipectl.s.physoftrst = 0;
+	/* Disable low power mode due to mcbuggin 29646 */
+	pipectl.s.suspend_en = 0;
+	write_uahc_reg(node, pipectl_addr, pipectl.u32);
+	udelay(100);
+	phycfg.u32 = read_uahc_reg(node, phycfg_addr);
+	phycfg.s.physoftrst = 0;
+	/* Disable low power mode due to mcbuggin 29646 */
+	phycfg.s.enblslpm = 0;
+	write_uahc_reg(node, phycfg_addr, phycfg.u32);
+
+	gctl.u32 = read_uahc_reg(node, gctl_addr);
+	gctl.s.coresoftreset = 0;
+	write_uahc_reg(node, gctl_addr, gctl.u32);
+	udelay(100);
+
+	/* Clear any previous bits set */
+	gsts.u32 = read_uahc_reg(node, gsts_addr);
+	write_uahc_reg(node, gsts_addr, gsts.u32);
+
+	/* We test that there hasn't been a PLL issue during
+	 * initialization. If we detect this then we retry.
+	 *
+	 * On octeon, the wce bit basically shouldn't do anything.  If
+	 * a PLL failed to come up then we will see a CSR timeout.
+	 */
+	portsc.u32 = read_uahc_reg(node, CVMX_USBDRDX_UAHC_PORTSCX(0, index));
+	portsc.s.wce = 1;
+	write_uahc_reg(node, CVMX_USBDRDX_UAHC_PORTSCX(0, index), portsc.u32);
+	gsts.u32 = read_uahc_reg(node, gsts_addr);
+	write_uahc_reg(node, gsts_addr, gsts.u32);
+	if (gsts.s.csrtimeout) {
+		dev_err(dev, "%s: Detected CSR tmout 0, can't continue...\n", __func__);
+		return -1;
+	}
+	udelay(100);
+	portsc.u32 = read_uahc_reg(node, CVMX_USBDRDX_UAHC_PORTSCX(1, index));
+	portsc.s.wce = 1;
+	write_uahc_reg(node, CVMX_USBDRDX_UAHC_PORTSCX(1, index), portsc.u32);
+	gsts.u32 = read_uahc_reg(node, gsts_addr);
+	write_uahc_reg(node, gsts_addr, gsts.u32);
+	if (gsts.s.csrtimeout) {
+		dev_err(dev, "%s: Detected CSR tmout 1, can't continue...\n", __func__);
+		return -1;
+	}
+
+	udelay(100);
+	portsc.u32 = read_uahc_reg(node, CVMX_USBDRDX_UAHC_PORTSCX(0, index));
+	portsc.s.wce = 0;
+	write_uahc_reg(node, CVMX_USBDRDX_UAHC_PORTSCX(0, index), portsc.u32);
+	gsts.u32 = read_uahc_reg(node, gsts_addr);
+	write_uahc_reg(node, gsts_addr, gsts.u32);
+	if (gsts.s.csrtimeout) {
+		dev_err(dev, "%s: Detected CSR tmout 2, can't continue...\n", __func__);
+		return -1;
+	}
+	udelay(100);
+	portsc.u32 = read_uahc_reg(node, CVMX_USBDRDX_UAHC_PORTSCX(0, index));
+	portsc.s.wce = 0;
+	write_uahc_reg(node, CVMX_USBDRDX_UAHC_PORTSCX(0, index), portsc.u32);
+	gsts.u32 = read_uahc_reg(node, gsts_addr);
+	write_uahc_reg(node, gsts_addr, gsts.u32);
+	if (gsts.s.csrtimeout) {
+		dev_err(dev, "%s: Detected CSR tmout 4, can't continue...\n", __func__);
+		return -1;
+	}
+
+	return 0;
+}
+
 static int __init dwc3_octeon_device_init(void)
 {
 	const char compat_node_name[] = "cavium,octeon-7130-usb-uctl";
@@ -531,6 +1117,12 @@ static int __init dwc3_octeon_device_init(void)
 				dev_info(&pdev->dev, "clocks initialized.\n");
 			dwc3_octeon_set_endian_mode((u64)base);
 			dwc3_octeon_phy_reset((u64)base);
+			if (dwc3_octeon_device_pll_lock(&pdev->dev, (u64)base) != 0) {
+				dev_err(&pdev->dev, "Failed to initialize clocks.");
+				mutex_unlock(&dwc3_octeon_clocks_mutex);
+				put_device(&pdev->dev);
+				return -EINVAL;
+			}
 			mutex_unlock(&dwc3_octeon_clocks_mutex);
 			devm_iounmap(&pdev->dev, base);
 			devm_release_mem_region(&pdev->dev, res->start,
Signature · 1 line
2.39.2
#1 lore raw

PATCH 1/11 MIPS: OCTEON: octeon-usb: add all register offsets

From: Ladislav Michl <[email protected]>

Glue code uses a mix of offset and absolute address register definition. Define all of them as offsets and use them consistently.

---
 arch/mips/cavium-octeon/octeon-usb.c | 35 +++++++++++++++++-----------
 1 file changed, 21 insertions(+), 14 deletions(-)
arch/mips/cavium-octeon/octeon-usb.c +21 -14
index 28677c615175..4d22eaa8a644 100644
@@ -17,6 +17,19 @@
 
 #include <asm/octeon/octeon.h>
 
+#define USBDRD_UCTL_CTL				0x00
+#define USBDRD_UCTL_BIST_STATUS			0x08
+#define USBDRD_UCTL_SPARE0			0x10
+#define USBDRD_UCTL_INTSTAT			0x30
+#define USBDRD_UCTL_PORT_CFG_HS(port)		(0x40 + (0x20 * port))
+#define USBDRD_UCTL_PORT_CFG_SS(port)		(0x48 + (0x20 * port))
+#define USBDRD_UCTL_PORT_CR_DBG_CFG(port)	(0x50 + (0x20 * port))
+#define USBDRD_UCTL_PORT_CR_DBG_STATUS(port)	(0x58 + (0x20 * port))
+#define USBDRD_UCTL_HOST_CFG			0xe0
+#define USBDRD_UCTL_SHIM_CFG			0xe8
+#define USBDRD_UCTL_ECC				0xf0
+#define USBDRD_UCTL_SPARE1			0xf8
+
 /* USB Control Register */
 union cvm_usbdrd_uctl_ctl {
 	uint64_t u64;
@@ -227,7 +240,6 @@ static uint8_t clk_div[OCTEON_H_CLKDIV_SEL] = {1, 2, 4, 6, 8, 16, 24, 32};
 
 static int dwc3_octeon_config_power(struct device *dev, u64 base)
 {
-#define UCTL_HOST_CFG	0xe0
 	union cvm_usbdrd_uctl_host_cfg uctl_host_cfg;
 	union cvmx_gpio_bit_cfgx gpio_bit;
 	uint32_t gpio_pwr[3];
@@ -268,16 +280,16 @@ static int dwc3_octeon_config_power(struct device *dev, u64 base)
 		}
 
 		/* Enable XHCI power control and set if active high or low. */
-		uctl_host_cfg.u64 = cvmx_read_csr(base + UCTL_HOST_CFG);
+		uctl_host_cfg.u64 = cvmx_read_csr(base + USBDRD_UCTL_HOST_CFG);
 		uctl_host_cfg.s.ppc_en = 1;
 		uctl_host_cfg.s.ppc_active_high_en = !power_active_low;
-		cvmx_write_csr(base + UCTL_HOST_CFG, uctl_host_cfg.u64);
+		cvmx_write_csr(base + USBDRD_UCTL_HOST_CFG, uctl_host_cfg.u64);
 	} else {
 		/* Disable XHCI power control and set if active high. */
-		uctl_host_cfg.u64 = cvmx_read_csr(base + UCTL_HOST_CFG);
+		uctl_host_cfg.u64 = cvmx_read_csr(base + USBDRD_UCTL_HOST_CFG);
 		uctl_host_cfg.s.ppc_en = 0;
 		uctl_host_cfg.s.ppc_active_high_en = 0;
-		cvmx_write_csr(base + UCTL_HOST_CFG, uctl_host_cfg.u64);
+		cvmx_write_csr(base + USBDRD_UCTL_HOST_CFG, uctl_host_cfg.u64);
 		dev_info(dev, "power control disabled\n");
 	}
 	return 0;
@@ -464,10 +476,9 @@ static int dwc3_octeon_clocks_start(struct device *dev, u64 base)
 
 static void __init dwc3_octeon_set_endian_mode(u64 base)
 {
-#define UCTL_SHIM_CFG	0xe8
 	union cvm_usbdrd_uctl_shim_cfg shim_cfg;
 
-	shim_cfg.u64 = cvmx_read_csr(base + UCTL_SHIM_CFG);
+	shim_cfg.u64 = cvmx_read_csr(base + USBDRD_UCTL_SHIM_CFG);
 #ifdef __BIG_ENDIAN
 	shim_cfg.s.dma_endian_mode = 1;
 	shim_cfg.s.csr_endian_mode = 1;
@@ -475,20 +486,16 @@ static void __init dwc3_octeon_set_endian_mode(u64 base)
 	shim_cfg.s.dma_endian_mode = 0;
 	shim_cfg.s.csr_endian_mode = 0;
 #endif
-	cvmx_write_csr(base + UCTL_SHIM_CFG, shim_cfg.u64);
+	cvmx_write_csr(base + USBDRD_UCTL_SHIM_CFG, shim_cfg.u64);
 }
 
-#define CVMX_USBDRDX_UCTL_CTL(index)				\
-		(CVMX_ADD_IO_SEG(0x0001180068000000ull) +	\
-		((index & 1) * 0x1000000ull))
 static void __init dwc3_octeon_phy_reset(u64 base)
 {
 	union cvm_usbdrd_uctl_ctl uctl_ctl;
-	int index = (base >> 24) & 1;
 
-	uctl_ctl.u64 = cvmx_read_csr(CVMX_USBDRDX_UCTL_CTL(index));
+	uctl_ctl.u64 = cvmx_read_csr(base + USBDRD_UCTL_CTL);
 	uctl_ctl.s.uphy_rst = 0;
-	cvmx_write_csr(CVMX_USBDRDX_UCTL_CTL(index), uctl_ctl.u64);
+	cvmx_write_csr(base + USBDRD_UCTL_CTL, uctl_ctl.u64);
 }
 
 static int __init dwc3_octeon_device_init(void)
Signature · 1 line
2.39.2
#2 lore raw

PATCH 2/11 MIPS: OCTEON: octeon-usb: use bitfields for control register

In reply to Ladislav Michl

From: Ladislav Michl <[email protected]>

Code needs to compile for all platforms in order to move it to drivers/usb/dwc3. Use Linux standard bitfield access macros to manipulate control register.

---
 arch/mips/cavium-octeon/octeon-usb.c | 330 +++++++++++++--------------
 1 file changed, 159 insertions(+), 171 deletions(-)
arch/mips/cavium-octeon/octeon-usb.c +159 -171
index 4d22eaa8a644..e8e57d0c3b14 100644
@@ -8,16 +8,117 @@
  * for more details.
  */
 
-#include <linux/module.h>
+#include <linux/bitfield.h>
+#include <linux/bits.h>
 #include <linux/device.h>
-#include <linux/mutex.h>
 #include <linux/delay.h>
-#include <linux/of_platform.h>
 #include <linux/io.h>
+#include <linux/module.h>
+#include <linux/mutex.h>
+#include <linux/of_platform.h>
 
 #include <asm/octeon/octeon.h>
 
+/*
+ * USB Control Register
+ */
 #define USBDRD_UCTL_CTL				0x00
+/* BIST fast-clear mode select. A BIST run with this bit set
+ * clears all entries in USBH RAMs to 0x0.
+ */
+# define USBDRD_UCTL_CTL_CLEAR_BIST		BIT(63)
+/* 1 = Start BIST and cleared by hardware */
+# define USBDRD_UCTL_CTL_START_BIST		BIT(62)
+/* Reference clock select for SuperSpeed and HighSpeed PLLs:
+ *	0x0 = Both PLLs use DLMC_REF_CLK0 for reference clock
+ *	0x1 = Both PLLs use DLMC_REF_CLK1 for reference clock
+ *	0x2 = SuperSpeed PLL uses DLMC_REF_CLK0 for reference clock &
+ *	      HighSpeed PLL uses PLL_REF_CLK for reference clck
+ *	0x3 = SuperSpeed PLL uses DLMC_REF_CLK1 for reference clock &
+ *	      HighSpeed PLL uses PLL_REF_CLK for reference clck
+ */
+# define USBDRD_UCTL_CTL_REF_CLK_SEL		GENMASK(61, 60)
+/* 1 = Spread-spectrum clock enable, 0 = SS clock disable */
+# define USBDRD_UCTL_CTL_SSC_EN			BIT(59)
+/* Spread-spectrum clock modulation range:
+ *	0x0 = -4980 ppm downspread
+ *	0x1 = -4492 ppm downspread
+ *	0x2 = -4003 ppm downspread
+ *	0x3 - 0x7 = Reserved
+ */
+# define USBDRD_UCTL_CTL_SSC_RANGE		GENMASK(58, 56)
+/* Enable non-standard oscillator frequencies:
+ *	[55:53] = modules -1
+ *	[52:47] = 2's complement push amount, 0 = Feature disabled
+ */
+# define USBDRD_UCTL_CTL_SSC_REF_CLK_SEL	GENMASK(55, 47)
+/* Reference clock multiplier for non-standard frequencies:
+ *	0x19 = 100MHz on DLMC_REF_CLK* if REF_CLK_SEL = 0x0 or 0x1
+ *	0x28 = 125MHz on DLMC_REF_CLK* if REF_CLK_SEL = 0x0 or 0x1
+ *	0x32 =  50MHz on DLMC_REF_CLK* if REF_CLK_SEL = 0x0 or 0x1
+ *	Other Values = Reserved
+ */
+# define USBDRD_UCTL_CTL_MPLL_MULTIPLIER	GENMASK(46, 40)
+/* Enable reference clock to prescaler for SuperSpeed functionality.
+ * Should always be set to "1"
+ */
+# define USBDRD_UCTL_CTL_REF_SSP_EN		BIT(39)
+/* Divide the reference clock by 2 before entering the
+ * REF_CLK_FSEL divider:
+ *	If REF_CLK_SEL = 0x0 or 0x1, then only 0x0 is legal
+ *	If REF_CLK_SEL = 0x2 or 0x3, then:
+ *		0x1 = DLMC_REF_CLK* is 125MHz
+ *		0x0 = DLMC_REF_CLK* is another supported frequency
+ */
+# define USBDRD_UCTL_CTL_REF_CLK_DIV2		BIT(38)
+/* Select reference clock freqnuency for both PLL blocks:
+ *	0x27 = REF_CLK_SEL is 0x0 or 0x1
+ *	0x07 = REF_CLK_SEL is 0x2 or 0x3
+ */
+# define USBDRD_UCTL_CTL_REF_CLK_FSEL		GENMASK(37, 32)
+/* Controller clock enable. */
+# define USBDRD_UCTL_CTL_H_CLK_EN		BIT(30)
+/* Select bypass input to controller clock divider:
+ *	0x0 = Use divided coprocessor clock from H_CLKDIV
+ *	0x1 = Use clock from GPIO pins
+ */
+# define USBDRD_UCTL_CTL_H_CLK_BYP_SEL		BIT(29)
+/* Reset controller clock divider. */
+# define USBDRD_UCTL_CTL_H_CLKDIV_RST		BIT(28)
+/* Clock divider select:
+ *	0x0 = divide by 1
+ *	0x1 = divide by 2
+ *	0x2 = divide by 4
+ *	0x3 = divide by 6
+ *	0x4 = divide by 8
+ *	0x5 = divide by 16
+ *	0x6 = divide by 24
+ *	0x7 = divide by 32
+ */
+# define USBDRD_UCTL_CTL_H_CLKDIV_SEL		GENMASK(26, 24)
+/* USB3 port permanently attached: 0x0 = No, 0x1 = Yes */
+# define USBDRD_UCTL_CTL_USB3_PORT_PERM_ATTACH	BIT(21)
+/* USB2 port permanently attached: 0x0 = No, 0x1 = Yes */
+# define USBDRD_UCTL_CTL_USB2_PORT_PERM_ATTACH	BIT(20)
+/* Disable SuperSpeed PHY: 0x0 = No, 0x1 = Yes */
+# define USBDRD_UCTL_CTL_USB3_PORT_DISABLE	BIT(18)
+/* Disable HighSpeed PHY: 0x0 = No, 0x1 = Yes */
+# define USBDRD_UCTL_CTL_USB2_PORT_DISABLE	BIT(16)
+/* Enable PHY SuperSpeed block power: 0x0 = No, 0x1 = Yes */
+# define USBDRD_UCTL_CTL_SS_POWER_EN		BIT(14)
+/* Enable PHY HighSpeed block power: 0x0 = No, 0x1 = Yes */
+# define USBDRD_UCTL_CTL_HS_POWER_EN		BIT(12)
+/* Enable USB UCTL interface clock: 0xx = No, 0x1 = Yes */
+# define USBDRD_UCTL_CTL_CSCLK_EN		BIT(4)
+/* Controller mode: 0x0 = Host, 0x1 = Device */
+# define USBDRD_UCTL_CTL_DRD_MODE		BIT(3)
+/* PHY reset */
+# define USBDRD_UCTL_CTL_UPHY_RST		BIT(2)
+/* Software reset UAHC */
+# define USBDRD_UCTL_CTL_UAHC_RST		BIT(1)
+/* Software resets UCTL */
+# define USBDRD_UCTL_CTL_UCTL_RST		BIT(0)
+
 #define USBDRD_UCTL_BIST_STATUS			0x08
 #define USBDRD_UCTL_SPARE0			0x10
 #define USBDRD_UCTL_INTSTAT			0x30
@@ -30,123 +131,6 @@
 #define USBDRD_UCTL_ECC				0xf0
 #define USBDRD_UCTL_SPARE1			0xf8
 
-/* USB Control Register */
-union cvm_usbdrd_uctl_ctl {
-	uint64_t u64;
-	struct cvm_usbdrd_uctl_ctl_s {
-	/* 1 = BIST and set all USB RAMs to 0x0, 0 = BIST */
-	__BITFIELD_FIELD(uint64_t clear_bist:1,
-	/* 1 = Start BIST and cleared by hardware */
-	__BITFIELD_FIELD(uint64_t start_bist:1,
-	/* Reference clock select for SuperSpeed and HighSpeed PLLs:
-	 *	0x0 = Both PLLs use DLMC_REF_CLK0 for reference clock
-	 *	0x1 = Both PLLs use DLMC_REF_CLK1 for reference clock
-	 *	0x2 = SuperSpeed PLL uses DLMC_REF_CLK0 for reference clock &
-	 *	      HighSpeed PLL uses PLL_REF_CLK for reference clck
-	 *	0x3 = SuperSpeed PLL uses DLMC_REF_CLK1 for reference clock &
-	 *	      HighSpeed PLL uses PLL_REF_CLK for reference clck
-	 */
-	__BITFIELD_FIELD(uint64_t ref_clk_sel:2,
-	/* 1 = Spread-spectrum clock enable, 0 = SS clock disable */
-	__BITFIELD_FIELD(uint64_t ssc_en:1,
-	/* Spread-spectrum clock modulation range:
-	 *	0x0 = -4980 ppm downspread
-	 *	0x1 = -4492 ppm downspread
-	 *	0x2 = -4003 ppm downspread
-	 *	0x3 - 0x7 = Reserved
-	 */
-	__BITFIELD_FIELD(uint64_t ssc_range:3,
-	/* Enable non-standard oscillator frequencies:
-	 *	[55:53] = modules -1
-	 *	[52:47] = 2's complement push amount, 0 = Feature disabled
-	 */
-	__BITFIELD_FIELD(uint64_t ssc_ref_clk_sel:9,
-	/* Reference clock multiplier for non-standard frequencies:
-	 *	0x19 = 100MHz on DLMC_REF_CLK* if REF_CLK_SEL = 0x0 or 0x1
-	 *	0x28 = 125MHz on DLMC_REF_CLK* if REF_CLK_SEL = 0x0 or 0x1
-	 *	0x32 =  50MHz on DLMC_REF_CLK* if REF_CLK_SEL = 0x0 or 0x1
-	 *	Other Values = Reserved
-	 */
-	__BITFIELD_FIELD(uint64_t mpll_multiplier:7,
-	/* Enable reference clock to prescaler for SuperSpeed functionality.
-	 * Should always be set to "1"
-	 */
-	__BITFIELD_FIELD(uint64_t ref_ssp_en:1,
-	/* Divide the reference clock by 2 before entering the
-	 * REF_CLK_FSEL divider:
-	 *	If REF_CLK_SEL = 0x0 or 0x1, then only 0x0 is legal
-	 *	If REF_CLK_SEL = 0x2 or 0x3, then:
-	 *		0x1 = DLMC_REF_CLK* is 125MHz
-	 *		0x0 = DLMC_REF_CLK* is another supported frequency
-	 */
-	__BITFIELD_FIELD(uint64_t ref_clk_div2:1,
-	/* Select reference clock freqnuency for both PLL blocks:
-	 *	0x27 = REF_CLK_SEL is 0x0 or 0x1
-	 *	0x07 = REF_CLK_SEL is 0x2 or 0x3
-	 */
-	__BITFIELD_FIELD(uint64_t ref_clk_fsel:6,
-	/* Reserved */
-	__BITFIELD_FIELD(uint64_t reserved_31_31:1,
-	/* Controller clock enable. */
-	__BITFIELD_FIELD(uint64_t h_clk_en:1,
-	/* Select bypass input to controller clock divider:
-	 *	0x0 = Use divided coprocessor clock from H_CLKDIV
-	 *	0x1 = Use clock from GPIO pins
-	 */
-	__BITFIELD_FIELD(uint64_t h_clk_byp_sel:1,
-	/* Reset controller clock divider. */
-	__BITFIELD_FIELD(uint64_t h_clkdiv_rst:1,
-	/* Reserved */
-	__BITFIELD_FIELD(uint64_t reserved_27_27:1,
-	/* Clock divider select:
-	 *	0x0 = divide by 1
-	 *	0x1 = divide by 2
-	 *	0x2 = divide by 4
-	 *	0x3 = divide by 6
-	 *	0x4 = divide by 8
-	 *	0x5 = divide by 16
-	 *	0x6 = divide by 24
-	 *	0x7 = divide by 32
-	 */
-	__BITFIELD_FIELD(uint64_t h_clkdiv_sel:3,
-	/* Reserved */
-	__BITFIELD_FIELD(uint64_t reserved_22_23:2,
-	/* USB3 port permanently attached: 0x0 = No, 0x1 = Yes */
-	__BITFIELD_FIELD(uint64_t usb3_port_perm_attach:1,
-	/* USB2 port permanently attached: 0x0 = No, 0x1 = Yes */
-	__BITFIELD_FIELD(uint64_t usb2_port_perm_attach:1,
-	/* Reserved */
-	__BITFIELD_FIELD(uint64_t reserved_19_19:1,
-	/* Disable SuperSpeed PHY: 0x0 = No, 0x1 = Yes */
-	__BITFIELD_FIELD(uint64_t usb3_port_disable:1,
-	/* Reserved */
-	__BITFIELD_FIELD(uint64_t reserved_17_17:1,
-	/* Disable HighSpeed PHY: 0x0 = No, 0x1 = Yes */
-	__BITFIELD_FIELD(uint64_t usb2_port_disable:1,
-	/* Reserved */
-	__BITFIELD_FIELD(uint64_t reserved_15_15:1,
-	/* Enable PHY SuperSpeed block power: 0x0 = No, 0x1 = Yes */
-	__BITFIELD_FIELD(uint64_t ss_power_en:1,
-	/* Reserved */
-	__BITFIELD_FIELD(uint64_t reserved_13_13:1,
-	/* Enable PHY HighSpeed block power: 0x0 = No, 0x1 = Yes */
-	__BITFIELD_FIELD(uint64_t hs_power_en:1,
-	/* Reserved */
-	__BITFIELD_FIELD(uint64_t reserved_5_11:7,
-	/* Enable USB UCTL interface clock: 0xx = No, 0x1 = Yes */
-	__BITFIELD_FIELD(uint64_t csclk_en:1,
-	/* Controller mode: 0x0 = Host, 0x1 = Device */
-	__BITFIELD_FIELD(uint64_t drd_mode:1,
-	/* PHY reset */
-	__BITFIELD_FIELD(uint64_t uphy_rst:1,
-	/* Software reset UAHC */
-	__BITFIELD_FIELD(uint64_t uahc_rst:1,
-	/* Software resets UCTL */
-	__BITFIELD_FIELD(uint64_t uctl_rst:1,
-	;)))))))))))))))))))))))))))))))))
-	} s;
-};
-
 /* UAHC Configuration Register */
 union cvm_usbdrd_uctl_host_cfg {
 	uint64_t u64;
@@ -297,14 +281,10 @@ static int dwc3_octeon_config_power(struct device *dev, u64 base)
 
 static int dwc3_octeon_clocks_start(struct device *dev, u64 base)
 {
-	union cvm_usbdrd_uctl_ctl uctl_ctl;
-	int ref_clk_sel = 2;
-	u64 div;
+	int i, mpll_mul, ref_clk_fsel, ref_clk_sel = 2;
 	u32 clock_rate;
-	int mpll_mul;
-	int i;
-	u64 h_clk_rate;
-	u64 uctl_ctl_reg = base;
+	u64 div, h_clk_rate, val;
+	u64 uctl_ctl_reg = base + USBDRD_UCTL_CTL;
 
 	if (dev->of_node) {
 		const char *ss_clock_type;
@@ -368,16 +348,16 @@ static int dwc3_octeon_clocks_start(struct device *dev, u64 base)
 	/* Step 2: Select GPIO for overcurrent indication, if desired. SKIP */
 
 	/* Step 3: Assert all resets. */
-	uctl_ctl.u64 = cvmx_read_csr(uctl_ctl_reg);
-	uctl_ctl.s.uphy_rst = 1;
-	uctl_ctl.s.uahc_rst = 1;
-	uctl_ctl.s.uctl_rst = 1;
-	cvmx_write_csr(uctl_ctl_reg, uctl_ctl.u64);
+	val = cvmx_read_csr(uctl_ctl_reg);
+	val |= USBDRD_UCTL_CTL_UPHY_RST |
+	       USBDRD_UCTL_CTL_UAHC_RST |
+	       USBDRD_UCTL_CTL_UCTL_RST;
+	cvmx_write_csr(uctl_ctl_reg, val);
 
 	/* Step 4a: Reset the clock dividers. */
-	uctl_ctl.u64 = cvmx_read_csr(uctl_ctl_reg);
-	uctl_ctl.s.h_clkdiv_rst = 1;
-	cvmx_write_csr(uctl_ctl_reg, uctl_ctl.u64);
+	val = cvmx_read_csr(uctl_ctl_reg);
+	val |= USBDRD_UCTL_CTL_H_CLKDIV_RST;
+	cvmx_write_csr(uctl_ctl_reg, val);
 
 	/* Step 4b: Select controller clock frequency. */
 	for (div = 0; div < OCTEON_H_CLKDIV_SEL; div++) {
@@ -386,26 +366,29 @@ static int dwc3_octeon_clocks_start(struct device *dev, u64 base)
 				 h_clk_rate >= OCTEON_MIN_H_CLK_RATE)
 			break;
 	}
-	uctl_ctl.u64 = cvmx_read_csr(uctl_ctl_reg);
-	uctl_ctl.s.h_clkdiv_sel = div;
-	uctl_ctl.s.h_clk_en = 1;
-	cvmx_write_csr(uctl_ctl_reg, uctl_ctl.u64);
-	uctl_ctl.u64 = cvmx_read_csr(uctl_ctl_reg);
-	if ((div != uctl_ctl.s.h_clkdiv_sel) || (!uctl_ctl.s.h_clk_en)) {
+	val = cvmx_read_csr(uctl_ctl_reg);
+	val &= ~USBDRD_UCTL_CTL_H_CLKDIV_SEL;
+	val |= FIELD_PREP(USBDRD_UCTL_CTL_H_CLKDIV_SEL, div);
+	val |= USBDRD_UCTL_CTL_H_CLK_EN;
+	cvmx_write_csr(uctl_ctl_reg, val);
+	val = cvmx_read_csr(uctl_ctl_reg);
+	if ((div != FIELD_GET(USBDRD_UCTL_CTL_H_CLKDIV_SEL, val)) ||
+	    (!(FIELD_GET(USBDRD_UCTL_CTL_H_CLK_EN, val)))) {
 		dev_err(dev, "dwc3 controller clock init failure.\n");
 			return -EINVAL;
 	}
 
 	/* Step 4c: Deassert the controller clock divider reset. */
-	uctl_ctl.u64 = cvmx_read_csr(uctl_ctl_reg);
-	uctl_ctl.s.h_clkdiv_rst = 0;
-	cvmx_write_csr(uctl_ctl_reg, uctl_ctl.u64);
+	val &= ~USBDRD_UCTL_CTL_H_CLKDIV_RST;
+	cvmx_write_csr(uctl_ctl_reg, val);
 
 	/* Step 5a: Reference clock configuration. */
-	uctl_ctl.u64 = cvmx_read_csr(uctl_ctl_reg);
-	uctl_ctl.s.ref_clk_sel = ref_clk_sel;
-	uctl_ctl.s.ref_clk_fsel = 0x07;
-	uctl_ctl.s.ref_clk_div2 = 0;
+	val = cvmx_read_csr(uctl_ctl_reg);
+	val &= ~USBDRD_UCTL_CTL_REF_CLK_DIV2;
+	val &= ~USBDRD_UCTL_CTL_REF_CLK_SEL;
+	val |= FIELD_PREP(USBDRD_UCTL_CTL_REF_CLK_SEL, ref_clk_sel);
+
+	ref_clk_fsel = 0x07;
 	switch (clock_rate) {
 	default:
 		dev_warn(dev, "Invalid ref_clk %u, using 100000000 instead\n",
@@ -414,7 +397,7 @@ static int dwc3_octeon_clocks_start(struct device *dev, u64 base)
 	case 100000000:
 		mpll_mul = 0x19;
 		if (ref_clk_sel < 2)
-			uctl_ctl.s.ref_clk_fsel = 0x27;
+			ref_clk_fsel = 0x27;
 		break;
 	case 50000000:
 		mpll_mul = 0x32;
@@ -423,28 +406,32 @@ static int dwc3_octeon_clocks_start(struct device *dev, u64 base)
 		mpll_mul = 0x28;
 		break;
 	}
-	uctl_ctl.s.mpll_multiplier = mpll_mul;
+	val &= ~USBDRD_UCTL_CTL_REF_CLK_FSEL;
+	val |= FIELD_PREP(USBDRD_UCTL_CTL_REF_CLK_FSEL, ref_clk_fsel);
+
+	val &= ~USBDRD_UCTL_CTL_MPLL_MULTIPLIER;
+	val |= FIELD_PREP(USBDRD_UCTL_CTL_MPLL_MULTIPLIER, mpll_mul);
 
 	/* Step 5b: Configure and enable spread-spectrum for SuperSpeed. */
-	uctl_ctl.s.ssc_en = 1;
+	val |= USBDRD_UCTL_CTL_SSC_EN;
 
 	/* Step 5c: Enable SuperSpeed. */
-	uctl_ctl.s.ref_ssp_en = 1;
+	val |= USBDRD_UCTL_CTL_REF_SSP_EN;
 
 	/* Step 5d: Configure PHYs. SKIP */
 
 	/* Step 6a & 6b: Power up PHYs. */
-	uctl_ctl.s.hs_power_en = 1;
-	uctl_ctl.s.ss_power_en = 1;
-	cvmx_write_csr(uctl_ctl_reg, uctl_ctl.u64);
+	val |= USBDRD_UCTL_CTL_HS_POWER_EN;
+	val |= USBDRD_UCTL_CTL_SS_POWER_EN;
+	cvmx_write_csr(uctl_ctl_reg, val);
 
 	/* Step 7: Wait 10 controller-clock cycles to take effect. */
 	udelay(10);
 
 	/* Step 8a: Deassert UCTL reset signal. */
-	uctl_ctl.u64 = cvmx_read_csr(uctl_ctl_reg);
-	uctl_ctl.s.uctl_rst = 0;
-	cvmx_write_csr(uctl_ctl_reg, uctl_ctl.u64);
+	val = cvmx_read_csr(uctl_ctl_reg);
+	val &= ~USBDRD_UCTL_CTL_UCTL_RST;
+	cvmx_write_csr(uctl_ctl_reg, val);
 
 	/* Step 8b: Wait 10 controller-clock cycles. */
 	udelay(10);
@@ -454,22 +441,22 @@ static int dwc3_octeon_clocks_start(struct device *dev, u64 base)
 		return -EINVAL;
 
 	/* Step 8d: Deassert UAHC reset signal. */
-	uctl_ctl.u64 = cvmx_read_csr(uctl_ctl_reg);
-	uctl_ctl.s.uahc_rst = 0;
-	cvmx_write_csr(uctl_ctl_reg, uctl_ctl.u64);
+	val = cvmx_read_csr(uctl_ctl_reg);
+	val &= ~USBDRD_UCTL_CTL_UAHC_RST;
+	cvmx_write_csr(uctl_ctl_reg, val);
 
 	/* Step 8e: Wait 10 controller-clock cycles. */
 	udelay(10);
 
 	/* Step 9: Enable conditional coprocessor clock of UCTL. */
-	uctl_ctl.u64 = cvmx_read_csr(uctl_ctl_reg);
-	uctl_ctl.s.csclk_en = 1;
-	cvmx_write_csr(uctl_ctl_reg, uctl_ctl.u64);
+	val = cvmx_read_csr(uctl_ctl_reg);
+	val |= USBDRD_UCTL_CTL_CSCLK_EN;
+	cvmx_write_csr(uctl_ctl_reg, val);
 
 	/*Step 10: Set for host mode only. */
-	uctl_ctl.u64 = cvmx_read_csr(uctl_ctl_reg);
-	uctl_ctl.s.drd_mode = 0;
-	cvmx_write_csr(uctl_ctl_reg, uctl_ctl.u64);
+	val = cvmx_read_csr(uctl_ctl_reg);
+	val &= ~USBDRD_UCTL_CTL_DRD_MODE;
+	cvmx_write_csr(uctl_ctl_reg, val);
 
 	return 0;
 }
@@ -491,11 +478,12 @@ static void __init dwc3_octeon_set_endian_mode(u64 base)
 
 static void __init dwc3_octeon_phy_reset(u64 base)
 {
-	union cvm_usbdrd_uctl_ctl uctl_ctl;
+	u64 val;
+	u64 uctl_ctl_reg = base + USBDRD_UCTL_CTL;
 
-	uctl_ctl.u64 = cvmx_read_csr(base + USBDRD_UCTL_CTL);
-	uctl_ctl.s.uphy_rst = 0;
-	cvmx_write_csr(base + USBDRD_UCTL_CTL, uctl_ctl.u64);
+	val = cvmx_read_csr(uctl_ctl_reg);
+	val &= ~USBDRD_UCTL_CTL_UPHY_RST;
+	cvmx_write_csr(uctl_ctl_reg, val);
 }
 
 static int __init dwc3_octeon_device_init(void)
Signature · 1 line
2.39.2
#3 lore raw

PATCH 3/11 MIPS: OCTEON: octeon-usb: use bitfields for host config register

In reply to Ladislav Michl

From: Ladislav Michl <[email protected]>

Use Linux standard bitfield access macros to manipulate host config register.

---
 arch/mips/cavium-octeon/octeon-usb.c | 83 +++++++++++++---------------
 1 file changed, 38 insertions(+), 45 deletions(-)
arch/mips/cavium-octeon/octeon-usb.c +38 -45
index e8e57d0c3b14..17051aee491d 100644
@@ -126,47 +126,36 @@
 #define USBDRD_UCTL_PORT_CFG_SS(port)		(0x48 + (0x20 * port))
 #define USBDRD_UCTL_PORT_CR_DBG_CFG(port)	(0x50 + (0x20 * port))
 #define USBDRD_UCTL_PORT_CR_DBG_STATUS(port)	(0x58 + (0x20 * port))
+
+/*
+ * UCTL Configuration Register
+ */
 #define USBDRD_UCTL_HOST_CFG			0xe0
+/* Indicates minimum value of all received BELT values */
+# define USBDRD_UCTL_HOST_CFG_HOST_CURRENT_BELT	GENMASK(59, 48)
+/* HS jitter adjustment */
+# define USBDRD_UCTL_HOST_CFG_FLA		GENMASK(37, 32)
+/* Bus-master enable: 0x0 = Disabled (stall DMAs), 0x1 = enabled */
+# define USBDRD_UCTL_HOST_CFG_BME		BIT(28)
+/* Overcurrent protection enable: 0x0 = unavailable, 0x1 = available */
+# define USBDRD_UCTL_HOST_OCI_EN		BIT(27)
+/* Overcurrent sene selection:
+ *	0x0 = Overcurrent indication from off-chip is active-low
+ *	0x1 = Overcurrent indication from off-chip is active-high
+ */
+# define USBDRD_UCTL_HOST_OCI_ACTIVE_HIGH_EN	BIT(26)
+/* Port power control enable: 0x0 = unavailable, 0x1 = available */
+# define USBDRD_UCTL_HOST_PPC_EN		BIT(25)
+/* Port power control sense selection:
+ *	0x0 = Port power to off-chip is active-low
+ *	0x1 = Port power to off-chip is active-high
+ */
+# define USBDRD_UCTL_HOST_PPC_ACTIVE_HIGH_EN	BIT(24)
+
 #define USBDRD_UCTL_SHIM_CFG			0xe8
 #define USBDRD_UCTL_ECC				0xf0
 #define USBDRD_UCTL_SPARE1			0xf8
 
-/* UAHC Configuration Register */
-union cvm_usbdrd_uctl_host_cfg {
-	uint64_t u64;
-	struct cvm_usbdrd_uctl_host_cfg_s {
-	/* Reserved */
-	__BITFIELD_FIELD(uint64_t reserved_60_63:4,
-	/* Indicates minimum value of all received BELT values */
-	__BITFIELD_FIELD(uint64_t host_current_belt:12,
-	/* Reserved */
-	__BITFIELD_FIELD(uint64_t reserved_38_47:10,
-	/* HS jitter adjustment */
-	__BITFIELD_FIELD(uint64_t fla:6,
-	/* Reserved */
-	__BITFIELD_FIELD(uint64_t reserved_29_31:3,
-	/* Bus-master enable: 0x0 = Disabled (stall DMAs), 0x1 = enabled */
-	__BITFIELD_FIELD(uint64_t bme:1,
-	/* Overcurrent protection enable: 0x0 = unavailable, 0x1 = available */
-	__BITFIELD_FIELD(uint64_t oci_en:1,
-	/* Overcurrent sene selection:
-	 *	0x0 = Overcurrent indication from off-chip is active-low
-	 *	0x1 = Overcurrent indication from off-chip is active-high
-	 */
-	__BITFIELD_FIELD(uint64_t oci_active_high_en:1,
-	/* Port power control enable: 0x0 = unavailable, 0x1 = available */
-	__BITFIELD_FIELD(uint64_t ppc_en:1,
-	/* Port power control sense selection:
-	 *	0x0 = Port power to off-chip is active-low
-	 *	0x1 = Port power to off-chip is active-high
-	 */
-	__BITFIELD_FIELD(uint64_t ppc_active_high_en:1,
-	/* Reserved */
-	__BITFIELD_FIELD(uint64_t reserved_0_23:24,
-	;)))))))))))
-	} s;
-};
-
 /* UCTL Shim Features Register */
 union cvm_usbdrd_uctl_shim_cfg {
 	uint64_t u64;
@@ -224,12 +213,13 @@ static uint8_t clk_div[OCTEON_H_CLKDIV_SEL] = {1, 2, 4, 6, 8, 16, 24, 32};
 
 static int dwc3_octeon_config_power(struct device *dev, u64 base)
 {
-	union cvm_usbdrd_uctl_host_cfg uctl_host_cfg;
 	union cvmx_gpio_bit_cfgx gpio_bit;
 	uint32_t gpio_pwr[3];
 	int gpio, len, power_active_low;
 	struct device_node *node = dev->of_node;
 	int index = (base >> 24) & 1;
+	u64 val;
+	u64 uctl_host_cfg_reg = base + USBDRD_UCTL_HOST_CFG;
 
 	if (of_find_property(node, "power", &len) != NULL) {
 		if (len == 12) {
@@ -264,16 +254,19 @@ static int dwc3_octeon_config_power(struct device *dev, u64 base)
 		}
 
 		/* Enable XHCI power control and set if active high or low. */
-		uctl_host_cfg.u64 = cvmx_read_csr(base + USBDRD_UCTL_HOST_CFG);
-		uctl_host_cfg.s.ppc_en = 1;
-		uctl_host_cfg.s.ppc_active_high_en = !power_active_low;
-		cvmx_write_csr(base + USBDRD_UCTL_HOST_CFG, uctl_host_cfg.u64);
+		val = cvmx_read_csr(uctl_host_cfg_reg);
+		val |= USBDRD_UCTL_HOST_PPC_EN;
+		if (power_active_low)
+			val &= ~USBDRD_UCTL_HOST_PPC_ACTIVE_HIGH_EN;
+		else
+			val |= USBDRD_UCTL_HOST_PPC_ACTIVE_HIGH_EN;
+		cvmx_write_csr(uctl_host_cfg_reg, val);
 	} else {
 		/* Disable XHCI power control and set if active high. */
-		uctl_host_cfg.u64 = cvmx_read_csr(base + USBDRD_UCTL_HOST_CFG);
-		uctl_host_cfg.s.ppc_en = 0;
-		uctl_host_cfg.s.ppc_active_high_en = 0;
-		cvmx_write_csr(base + USBDRD_UCTL_HOST_CFG, uctl_host_cfg.u64);
+		val = cvmx_read_csr(uctl_host_cfg_reg);
+		val &= ~USBDRD_UCTL_HOST_PPC_EN;
+		val &= ~USBDRD_UCTL_HOST_PPC_ACTIVE_HIGH_EN;
+		cvmx_write_csr(uctl_host_cfg_reg, val);
 		dev_info(dev, "power control disabled\n");
 	}
 	return 0;
Signature · 1 line
2.39.2
#4 lore raw

PATCH 4/11 MIPS: OCTEON: octeon-usb: use bitfields for shim register

In reply to Ladislav Michl

From: Ladislav Michl <[email protected]>

Use Linux standard bitfield access macros to manipulate shim register.

---
 arch/mips/cavium-octeon/octeon-usb.c | 96 ++++++++++++----------------
 1 file changed, 41 insertions(+), 55 deletions(-)
arch/mips/cavium-octeon/octeon-usb.c +41 -55
index 17051aee491d..1c48ee77125a 100644
@@ -152,57 +152,43 @@
  */
 # define USBDRD_UCTL_HOST_PPC_ACTIVE_HIGH_EN	BIT(24)
 
+/*
+ * UCTL Shim Features Register
+ */
 #define USBDRD_UCTL_SHIM_CFG			0xe8
+/* Out-of-bound UAHC register access: 0 = read, 1 = write */
+# define USBDRD_UCTL_SHIM_CFG_XS_NCB_OOB_WRN	BIT(63)
+/* SRCID error log for out-of-bound UAHC register access:
+ *	[59:58] = chipID
+ *	[57] = Request source: 0 = core, 1 = NCB-device
+ *	[56:51] = Core/NCB-device number, [56] always 0 for NCB devices
+ *	[50:48] = SubID
+ */
+# define USBDRD_UCTL_SHIM_CFG_XS_NCB_OOB_OSRC	GENMASK(59, 48)
+/* Error log for bad UAHC DMA access: 0 = Read log, 1 = Write log */
+# define USBDRD_UCTL_SHIM_CFG_XM_BAD_DMA_WRN	BIT(47)
+/* Encoded error type for bad UAHC DMA */
+# define USBDRD_UCTL_SHIM_CFG_XM_BAD_DMA_TYPE	GENMASK(43, 40)
+/* Select the IOI read command used by DMA accesses */
+# define USBDRD_UCTL_SHIM_CFG_DMA_READ_CMD	BIT(12)
+/* Select endian format for DMA accesses to the L2C:
+ *	0x0 = Little endian
+ *	0x1 = Big endian
+ *	0x2 = Reserved
+ *	0x3 = Reserved
+ */
+# define USBDRD_UCTL_SHIM_CFG_DMA_ENDIAN_MODE	GENMASK(9, 8)
+/* Select endian format for IOI CSR access to UAHC:
+ *	0x0 = Little endian
+ *	0x1 = Big endian
+ *	0x2 = Reserved
+ *	0x3 = Reserved
+ */
+# define USBDRD_UCTL_SHIM_CFG_CSR_ENDIAN_MODE	GENMASK(1, 0)
+
 #define USBDRD_UCTL_ECC				0xf0
 #define USBDRD_UCTL_SPARE1			0xf8
 
-/* UCTL Shim Features Register */
-union cvm_usbdrd_uctl_shim_cfg {
-	uint64_t u64;
-	struct cvm_usbdrd_uctl_shim_cfg_s {
-	/* Out-of-bound UAHC register access: 0 = read, 1 = write */
-	__BITFIELD_FIELD(uint64_t xs_ncb_oob_wrn:1,
-	/* Reserved */
-	__BITFIELD_FIELD(uint64_t reserved_60_62:3,
-	/* SRCID error log for out-of-bound UAHC register access:
-	 *	[59:58] = chipID
-	 *	[57] = Request source: 0 = core, 1 = NCB-device
-	 *	[56:51] = Core/NCB-device number, [56] always 0 for NCB devices
-	 *	[50:48] = SubID
-	 */
-	__BITFIELD_FIELD(uint64_t xs_ncb_oob_osrc:12,
-	/* Error log for bad UAHC DMA access: 0 = Read log, 1 = Write log */
-	__BITFIELD_FIELD(uint64_t xm_bad_dma_wrn:1,
-	/* Reserved */
-	__BITFIELD_FIELD(uint64_t reserved_44_46:3,
-	/* Encoded error type for bad UAHC DMA */
-	__BITFIELD_FIELD(uint64_t xm_bad_dma_type:4,
-	/* Reserved */
-	__BITFIELD_FIELD(uint64_t reserved_13_39:27,
-	/* Select the IOI read command used by DMA accesses */
-	__BITFIELD_FIELD(uint64_t dma_read_cmd:1,
-	/* Reserved */
-	__BITFIELD_FIELD(uint64_t reserved_10_11:2,
-	/* Select endian format for DMA accesses to the L2c:
-	 *	0x0 = Little endian
-	 *`	0x1 = Big endian
-	 *	0x2 = Reserved
-	 *	0x3 = Reserved
-	 */
-	__BITFIELD_FIELD(uint64_t dma_endian_mode:2,
-	/* Reserved */
-	__BITFIELD_FIELD(uint64_t reserved_2_7:6,
-	/* Select endian format for IOI CSR access to UAHC:
-	 *	0x0 = Little endian
-	 *`	0x1 = Big endian
-	 *	0x2 = Reserved
-	 *	0x3 = Reserved
-	 */
-	__BITFIELD_FIELD(uint64_t csr_endian_mode:2,
-	;))))))))))))
-	} s;
-};
-
 #define OCTEON_H_CLKDIV_SEL		8
 #define OCTEON_MIN_H_CLK_RATE		150000000
 #define OCTEON_MAX_H_CLK_RATE		300000000
@@ -456,17 +442,17 @@ static int dwc3_octeon_clocks_start(struct device *dev, u64 base)
 
 static void __init dwc3_octeon_set_endian_mode(u64 base)
 {
-	union cvm_usbdrd_uctl_shim_cfg shim_cfg;
+	u64 val;
+	u64 uctl_shim_cfg_reg = base + USBDRD_UCTL_SHIM_CFG;
 
-	shim_cfg.u64 = cvmx_read_csr(base + USBDRD_UCTL_SHIM_CFG);
+	val = cvmx_read_csr(uctl_shim_cfg_reg);
+	val &= ~USBDRD_UCTL_SHIM_CFG_DMA_ENDIAN_MODE;
+	val &= ~USBDRD_UCTL_SHIM_CFG_CSR_ENDIAN_MODE;
 #ifdef __BIG_ENDIAN
-	shim_cfg.s.dma_endian_mode = 1;
-	shim_cfg.s.csr_endian_mode = 1;
-#else
-	shim_cfg.s.dma_endian_mode = 0;
-	shim_cfg.s.csr_endian_mode = 0;
+	val |= FIELD_PREP(USBDRD_UCTL_SHIM_CFG_DMA_ENDIAN_MODE, 1);
+	val |= FIELD_PREP(USBDRD_UCTL_SHIM_CFG_CSR_ENDIAN_MODE, 1);
 #endif
-	cvmx_write_csr(base + USBDRD_UCTL_SHIM_CFG, shim_cfg.u64);
+	cvmx_write_csr(uctl_shim_cfg_reg, val);
 }
 
 static void __init dwc3_octeon_phy_reset(u64 base)
Signature · 1 line
2.39.2
#5 lore raw

PATCH 5/11 MIPS: OCTEON: octeon-usb: move gpio config to separate function

In reply to Ladislav Michl

Power gpio configuration is using Octeon specific code, so move it to separate function, that can later be guarded with ifdefs.

---
 arch/mips/cavium-octeon/octeon-usb.c | 45 +++++++++++++++-------------
 1 file changed, 25 insertions(+), 20 deletions(-)
arch/mips/cavium-octeon/octeon-usb.c +25 -20
index 1c48ee77125a..0f9800b3d373 100644
@@ -197,13 +197,35 @@ static DEFINE_MUTEX(dwc3_octeon_clocks_mutex);
 static uint8_t clk_div[OCTEON_H_CLKDIV_SEL] = {1, 2, 4, 6, 8, 16, 24, 32};
 
 
-static int dwc3_octeon_config_power(struct device *dev, u64 base)
+static void dwc3_octeon_config_gpio(int index, int gpio)
 {
 	union cvmx_gpio_bit_cfgx gpio_bit;
+
+	if ((OCTEON_IS_MODEL(OCTEON_CN73XX) ||
+	    OCTEON_IS_MODEL(OCTEON_CNF75XX))
+	    && gpio <= 31) {
+		gpio_bit.u64 = cvmx_read_csr(CVMX_GPIO_BIT_CFGX(gpio));
+		gpio_bit.s.tx_oe = 1;
+		gpio_bit.s.output_sel = (index == 0 ? 0x14 : 0x15);
+		cvmx_write_csr(CVMX_GPIO_BIT_CFGX(gpio), gpio_bit.u64);
+	} else if (gpio <= 15) {
+		gpio_bit.u64 = cvmx_read_csr(CVMX_GPIO_BIT_CFGX(gpio));
+		gpio_bit.s.tx_oe = 1;
+		gpio_bit.s.output_sel = (index == 0 ? 0x14 : 0x19);
+		cvmx_write_csr(CVMX_GPIO_BIT_CFGX(gpio), gpio_bit.u64);
+	} else {
+		gpio_bit.u64 = cvmx_read_csr(CVMX_GPIO_XBIT_CFGX(gpio));
+		gpio_bit.s.tx_oe = 1;
+		gpio_bit.s.output_sel = (index == 0 ? 0x14 : 0x19);
+		cvmx_write_csr(CVMX_GPIO_XBIT_CFGX(gpio), gpio_bit.u64);
+	}
+}
+
+static int dwc3_octeon_config_power(struct device *dev, u64 base)
+{
 	uint32_t gpio_pwr[3];
 	int gpio, len, power_active_low;
 	struct device_node *node = dev->of_node;
-	int index = (base >> 24) & 1;
 	u64 val;
 	u64 uctl_host_cfg_reg = base + USBDRD_UCTL_HOST_CFG;
 
@@ -220,24 +242,7 @@ static int dwc3_octeon_config_power(struct device *dev, u64 base)
 			dev_err(dev, "invalid power configuration\n");
 			return -EINVAL;
 		}
-		if ((OCTEON_IS_MODEL(OCTEON_CN73XX) ||
-		    OCTEON_IS_MODEL(OCTEON_CNF75XX))
-		    && gpio <= 31) {
-			gpio_bit.u64 = cvmx_read_csr(CVMX_GPIO_BIT_CFGX(gpio));
-			gpio_bit.s.tx_oe = 1;
-			gpio_bit.s.output_sel = (index == 0 ? 0x14 : 0x15);
-			cvmx_write_csr(CVMX_GPIO_BIT_CFGX(gpio), gpio_bit.u64);
-		} else if (gpio <= 15) {
-			gpio_bit.u64 = cvmx_read_csr(CVMX_GPIO_BIT_CFGX(gpio));
-			gpio_bit.s.tx_oe = 1;
-			gpio_bit.s.output_sel = (index == 0 ? 0x14 : 0x19);
-			cvmx_write_csr(CVMX_GPIO_BIT_CFGX(gpio), gpio_bit.u64);
-		} else {
-			gpio_bit.u64 = cvmx_read_csr(CVMX_GPIO_XBIT_CFGX(gpio));
-			gpio_bit.s.tx_oe = 1;
-			gpio_bit.s.output_sel = (index == 0 ? 0x14 : 0x19);
-			cvmx_write_csr(CVMX_GPIO_XBIT_CFGX(gpio), gpio_bit.u64);
-		}
+		dwc3_octeon_config_gpio((base >> 24) & 1, gpio);
 
 		/* Enable XHCI power control and set if active high or low. */
 		val = cvmx_read_csr(uctl_host_cfg_reg);
Signature · 1 line
2.39.2
#6 lore raw

PATCH 6/11 MIPS: OCTEON: octeon-usb: introduce dwc3_octeon_{read,write}q

In reply to Ladislav Michl

From: Ladislav Michl <[email protected]>

Move all register access code into separate functions and provide their no-op version for non Octeon platforms.

Later it might be possible to replace them with standard Linux functions, however datasheets are not publicly available and I have only one Octeon board to test, so lets stay on safe side for now.

---
 arch/mips/cavium-octeon/octeon-usb.c | 99 +++++++++++++++++-----------
 1 file changed, 59 insertions(+), 40 deletions(-)
arch/mips/cavium-octeon/octeon-usb.c +59 -40
index 0f9800b3d373..aaddc874f0ca 100644
@@ -17,8 +17,6 @@
 #include <linux/mutex.h>
 #include <linux/of_platform.h>
 
-#include <asm/octeon/octeon.h>
-
 /*
  * USB Control Register
  */
@@ -196,6 +194,17 @@
 static DEFINE_MUTEX(dwc3_octeon_clocks_mutex);
 static uint8_t clk_div[OCTEON_H_CLKDIV_SEL] = {1, 2, 4, 6, 8, 16, 24, 32};
 
+#ifdef CONFIG_CAVIUM_OCTEON_SOC
+#include <asm/octeon/octeon.h>
+static inline uint64_t dwc3_octeon_readq(void __iomem *addr)
+{
+	return cvmx_readq_csr(addr);
+}
+
+static inline void dwc3_octeon_writeq(void __iomem *base, uint64_t val)
+{
+	cvmx_writeq_csr(base, val);
+}
 
 static void dwc3_octeon_config_gpio(int index, int gpio)
 {
@@ -220,14 +229,24 @@ static void dwc3_octeon_config_gpio(int index, int gpio)
 		cvmx_write_csr(CVMX_GPIO_XBIT_CFGX(gpio), gpio_bit.u64);
 	}
 }
+#else
+static inline uint64_t dwc3_octeon_readq(void __iomem *addr)
+{
+	return 0;
+}
+
+static inline void dwc3_octeon_writeq(void __iomem *base, uint64_t val) { }
+
+static inline void dwc3_octeon_config_gpio(int index, int gpio) { }
+#endif
 
-static int dwc3_octeon_config_power(struct device *dev, u64 base)
+static int dwc3_octeon_config_power(struct device *dev, void __iomem *base)
 {
 	uint32_t gpio_pwr[3];
 	int gpio, len, power_active_low;
 	struct device_node *node = dev->of_node;
 	u64 val;
-	u64 uctl_host_cfg_reg = base + USBDRD_UCTL_HOST_CFG;
+	void __iomem *uctl_host_cfg_reg = base + USBDRD_UCTL_HOST_CFG;
 
 	if (of_find_property(node, "power", &len) != NULL) {
 		if (len == 12) {
@@ -242,33 +261,33 @@ static int dwc3_octeon_config_power(struct device *dev, u64 base)
 			dev_err(dev, "invalid power configuration\n");
 			return -EINVAL;
 		}
-		dwc3_octeon_config_gpio((base >> 24) & 1, gpio);
+		dwc3_octeon_config_gpio(((u64)base >> 24) & 1, gpio);
 
 		/* Enable XHCI power control and set if active high or low. */
-		val = cvmx_read_csr(uctl_host_cfg_reg);
+		val = dwc3_octeon_readq(uctl_host_cfg_reg);
 		val |= USBDRD_UCTL_HOST_PPC_EN;
 		if (power_active_low)
 			val &= ~USBDRD_UCTL_HOST_PPC_ACTIVE_HIGH_EN;
 		else
 			val |= USBDRD_UCTL_HOST_PPC_ACTIVE_HIGH_EN;
-		cvmx_write_csr(uctl_host_cfg_reg, val);
+		dwc3_octeon_writeq(uctl_host_cfg_reg, val);
 	} else {
 		/* Disable XHCI power control and set if active high. */
-		val = cvmx_read_csr(uctl_host_cfg_reg);
+		val = dwc3_octeon_readq(uctl_host_cfg_reg);
 		val &= ~USBDRD_UCTL_HOST_PPC_EN;
 		val &= ~USBDRD_UCTL_HOST_PPC_ACTIVE_HIGH_EN;
-		cvmx_write_csr(uctl_host_cfg_reg, val);
+		dwc3_octeon_writeq(uctl_host_cfg_reg, val);
 		dev_info(dev, "power control disabled\n");
 	}
 	return 0;
 }
 
-static int dwc3_octeon_clocks_start(struct device *dev, u64 base)
+static int dwc3_octeon_clocks_start(struct device *dev, void __iomem *base)
 {
 	int i, mpll_mul, ref_clk_fsel, ref_clk_sel = 2;
 	u32 clock_rate;
 	u64 div, h_clk_rate, val;
-	u64 uctl_ctl_reg = base + USBDRD_UCTL_CTL;
+	void __iomem *uctl_ctl_reg = base + USBDRD_UCTL_CTL;
 
 	if (dev->of_node) {
 		const char *ss_clock_type;
@@ -332,16 +351,16 @@ static int dwc3_octeon_clocks_start(struct device *dev, u64 base)
 	/* Step 2: Select GPIO for overcurrent indication, if desired. SKIP */
 
 	/* Step 3: Assert all resets. */
-	val = cvmx_read_csr(uctl_ctl_reg);
+	val = dwc3_octeon_readq(uctl_ctl_reg);
 	val |= USBDRD_UCTL_CTL_UPHY_RST |
 	       USBDRD_UCTL_CTL_UAHC_RST |
 	       USBDRD_UCTL_CTL_UCTL_RST;
-	cvmx_write_csr(uctl_ctl_reg, val);
+	dwc3_octeon_writeq(uctl_ctl_reg, val);
 
 	/* Step 4a: Reset the clock dividers. */
-	val = cvmx_read_csr(uctl_ctl_reg);
+	val = dwc3_octeon_readq(uctl_ctl_reg);
 	val |= USBDRD_UCTL_CTL_H_CLKDIV_RST;
-	cvmx_write_csr(uctl_ctl_reg, val);
+	dwc3_octeon_writeq(uctl_ctl_reg, val);
 
 	/* Step 4b: Select controller clock frequency. */
 	for (div = 0; div < OCTEON_H_CLKDIV_SEL; div++) {
@@ -350,12 +369,12 @@ static int dwc3_octeon_clocks_start(struct device *dev, u64 base)
 				 h_clk_rate >= OCTEON_MIN_H_CLK_RATE)
 			break;
 	}
-	val = cvmx_read_csr(uctl_ctl_reg);
+	val = dwc3_octeon_readq(uctl_ctl_reg);
 	val &= ~USBDRD_UCTL_CTL_H_CLKDIV_SEL;
 	val |= FIELD_PREP(USBDRD_UCTL_CTL_H_CLKDIV_SEL, div);
 	val |= USBDRD_UCTL_CTL_H_CLK_EN;
-	cvmx_write_csr(uctl_ctl_reg, val);
-	val = cvmx_read_csr(uctl_ctl_reg);
+	dwc3_octeon_writeq(uctl_ctl_reg, val);
+	val = dwc3_octeon_readq(uctl_ctl_reg);
 	if ((div != FIELD_GET(USBDRD_UCTL_CTL_H_CLKDIV_SEL, val)) ||
 	    (!(FIELD_GET(USBDRD_UCTL_CTL_H_CLK_EN, val)))) {
 		dev_err(dev, "dwc3 controller clock init failure.\n");
@@ -364,10 +383,10 @@ static int dwc3_octeon_clocks_start(struct device *dev, u64 base)
 
 	/* Step 4c: Deassert the controller clock divider reset. */
 	val &= ~USBDRD_UCTL_CTL_H_CLKDIV_RST;
-	cvmx_write_csr(uctl_ctl_reg, val);
+	dwc3_octeon_writeq(uctl_ctl_reg, val);
 
 	/* Step 5a: Reference clock configuration. */
-	val = cvmx_read_csr(uctl_ctl_reg);
+	val = dwc3_octeon_readq(uctl_ctl_reg);
 	val &= ~USBDRD_UCTL_CTL_REF_CLK_DIV2;
 	val &= ~USBDRD_UCTL_CTL_REF_CLK_SEL;
 	val |= FIELD_PREP(USBDRD_UCTL_CTL_REF_CLK_SEL, ref_clk_sel);
@@ -407,15 +426,15 @@ static int dwc3_octeon_clocks_start(struct device *dev, u64 base)
 	/* Step 6a & 6b: Power up PHYs. */
 	val |= USBDRD_UCTL_CTL_HS_POWER_EN;
 	val |= USBDRD_UCTL_CTL_SS_POWER_EN;
-	cvmx_write_csr(uctl_ctl_reg, val);
+	dwc3_octeon_writeq(uctl_ctl_reg, val);
 
 	/* Step 7: Wait 10 controller-clock cycles to take effect. */
 	udelay(10);
 
 	/* Step 8a: Deassert UCTL reset signal. */
-	val = cvmx_read_csr(uctl_ctl_reg);
+	val = dwc3_octeon_readq(uctl_ctl_reg);
 	val &= ~USBDRD_UCTL_CTL_UCTL_RST;
-	cvmx_write_csr(uctl_ctl_reg, val);
+	dwc3_octeon_writeq(uctl_ctl_reg, val);
 
 	/* Step 8b: Wait 10 controller-clock cycles. */
 	udelay(10);
@@ -425,49 +444,49 @@ static int dwc3_octeon_clocks_start(struct device *dev, u64 base)
 		return -EINVAL;
 
 	/* Step 8d: Deassert UAHC reset signal. */
-	val = cvmx_read_csr(uctl_ctl_reg);
+	val = dwc3_octeon_readq(uctl_ctl_reg);
 	val &= ~USBDRD_UCTL_CTL_UAHC_RST;
-	cvmx_write_csr(uctl_ctl_reg, val);
+	dwc3_octeon_writeq(uctl_ctl_reg, val);
 
 	/* Step 8e: Wait 10 controller-clock cycles. */
 	udelay(10);
 
 	/* Step 9: Enable conditional coprocessor clock of UCTL. */
-	val = cvmx_read_csr(uctl_ctl_reg);
+	val = dwc3_octeon_readq(uctl_ctl_reg);
 	val |= USBDRD_UCTL_CTL_CSCLK_EN;
-	cvmx_write_csr(uctl_ctl_reg, val);
+	dwc3_octeon_writeq(uctl_ctl_reg, val);
 
 	/*Step 10: Set for host mode only. */
-	val = cvmx_read_csr(uctl_ctl_reg);
+	val = dwc3_octeon_readq(uctl_ctl_reg);
 	val &= ~USBDRD_UCTL_CTL_DRD_MODE;
-	cvmx_write_csr(uctl_ctl_reg, val);
+	dwc3_octeon_writeq(uctl_ctl_reg, val);
 
 	return 0;
 }
 
-static void __init dwc3_octeon_set_endian_mode(u64 base)
+static void __init dwc3_octeon_set_endian_mode(void __iomem *base)
 {
 	u64 val;
-	u64 uctl_shim_cfg_reg = base + USBDRD_UCTL_SHIM_CFG;
+	void __iomem *uctl_shim_cfg_reg = base + USBDRD_UCTL_SHIM_CFG;
 
-	val = cvmx_read_csr(uctl_shim_cfg_reg);
+	val = dwc3_octeon_readq(uctl_shim_cfg_reg);
 	val &= ~USBDRD_UCTL_SHIM_CFG_DMA_ENDIAN_MODE;
 	val &= ~USBDRD_UCTL_SHIM_CFG_CSR_ENDIAN_MODE;
 #ifdef __BIG_ENDIAN
 	val |= FIELD_PREP(USBDRD_UCTL_SHIM_CFG_DMA_ENDIAN_MODE, 1);
 	val |= FIELD_PREP(USBDRD_UCTL_SHIM_CFG_CSR_ENDIAN_MODE, 1);
 #endif
-	cvmx_write_csr(uctl_shim_cfg_reg, val);
+	dwc3_octeon_writeq(uctl_shim_cfg_reg, val);
 }
 
-static void __init dwc3_octeon_phy_reset(u64 base)
+static void __init dwc3_octeon_phy_reset(void __iomem *base)
 {
 	u64 val;
-	u64 uctl_ctl_reg = base + USBDRD_UCTL_CTL;
+	void __iomem *uctl_ctl_reg = base + USBDRD_UCTL_CTL;
 
-	val = cvmx_read_csr(uctl_ctl_reg);
+	val = dwc3_octeon_readq(uctl_ctl_reg);
 	val &= ~USBDRD_UCTL_CTL_UPHY_RST;
-	cvmx_write_csr(uctl_ctl_reg, val);
+	dwc3_octeon_writeq(uctl_ctl_reg, val);
 }
 
 static int __init dwc3_octeon_device_init(void)
@@ -506,10 +525,10 @@ static int __init dwc3_octeon_device_init(void)
 			}
 
 			mutex_lock(&dwc3_octeon_clocks_mutex);
-			if (dwc3_octeon_clocks_start(&pdev->dev, (u64)base) == 0)
+			if (dwc3_octeon_clocks_start(&pdev->dev, base) == 0)
 				dev_info(&pdev->dev, "clocks initialized.\n");
-			dwc3_octeon_set_endian_mode((u64)base);
-			dwc3_octeon_phy_reset((u64)base);
+			dwc3_octeon_set_endian_mode(base);
+			dwc3_octeon_phy_reset(base);
 			mutex_unlock(&dwc3_octeon_clocks_mutex);
 			devm_iounmap(&pdev->dev, base);
 			devm_release_mem_region(&pdev->dev, res->start,
Signature · 1 line
2.39.2
#7 lore raw

PATCH 7/11 MIPS: OCTEON: octeon-usb: cleanup divider calculation

In reply to Ladislav Michl

From: Ladislav Michl <[email protected]>

Simple self-contained function is easier to review.

---
 arch/mips/cavium-octeon/octeon-usb.c | 31 ++++++++++++++++------------
 1 file changed, 18 insertions(+), 13 deletions(-)
arch/mips/cavium-octeon/octeon-usb.c +18 -13
index aaddc874f0ca..2add435ad038 100644
@@ -187,12 +187,7 @@
 #define USBDRD_UCTL_ECC				0xf0
 #define USBDRD_UCTL_SPARE1			0xf8
 
-#define OCTEON_H_CLKDIV_SEL		8
-#define OCTEON_MIN_H_CLK_RATE		150000000
-#define OCTEON_MAX_H_CLK_RATE		300000000
-
 static DEFINE_MUTEX(dwc3_octeon_clocks_mutex);
-static uint8_t clk_div[OCTEON_H_CLKDIV_SEL] = {1, 2, 4, 6, 8, 16, 24, 32};
 
 #ifdef CONFIG_CAVIUM_OCTEON_SOC
 #include <asm/octeon/octeon.h>
@@ -240,6 +235,21 @@ static inline void dwc3_octeon_writeq(void __iomem *base, uint64_t val) { }
 static inline void dwc3_octeon_config_gpio(int index, int gpio) { }
 #endif
 
+static int dwc3_octeon_get_divider(void)
+{
+	static const uint8_t clk_div[] = { 1, 2, 4, 6, 8, 16, 24, 32 };
+	int div = 0;
+
+	while (div < ARRAY_SIZE(clk_div)) {
+		uint64_t rate = octeon_get_io_clock_rate() / clk_div[div];
+		if (rate <= 300000000 && rate >= 150000000)
+			break;
+		div++;
+	}
+
+	return div;
+}
+
 static int dwc3_octeon_config_power(struct device *dev, void __iomem *base)
 {
 	uint32_t gpio_pwr[3];
@@ -284,9 +294,9 @@ static int dwc3_octeon_config_power(struct device *dev, void __iomem *base)
 
 static int dwc3_octeon_clocks_start(struct device *dev, void __iomem *base)
 {
-	int i, mpll_mul, ref_clk_fsel, ref_clk_sel = 2;
+	int i, div, mpll_mul, ref_clk_fsel, ref_clk_sel = 2;
 	u32 clock_rate;
-	u64 div, h_clk_rate, val;
+	u64 val;
 	void __iomem *uctl_ctl_reg = base + USBDRD_UCTL_CTL;
 
 	if (dev->of_node) {
@@ -363,12 +373,7 @@ static int dwc3_octeon_clocks_start(struct device *dev, void __iomem *base)
 	dwc3_octeon_writeq(uctl_ctl_reg, val);
 
 	/* Step 4b: Select controller clock frequency. */
-	for (div = 0; div < OCTEON_H_CLKDIV_SEL; div++) {
-		h_clk_rate = octeon_get_io_clock_rate() / clk_div[div];
-		if (h_clk_rate <= OCTEON_MAX_H_CLK_RATE &&
-				 h_clk_rate >= OCTEON_MIN_H_CLK_RATE)
-			break;
-	}
+	div = dwc3_octeon_get_divider();
 	val = dwc3_octeon_readq(uctl_ctl_reg);
 	val &= ~USBDRD_UCTL_CTL_H_CLKDIV_SEL;
 	val |= FIELD_PREP(USBDRD_UCTL_CTL_H_CLKDIV_SEL, div);
Signature · 1 line
2.39.2
#8 lore raw

PATCH 8/11 usb: dwc3: Move Octeon glue code from arch/mips

In reply to Ladislav Michl

From: Ladislav Michl <[email protected]>

Octeon DWC3 glue code now compiles on all platforms, so move it to drivers/usb/dwc3. No functional changes.

---
 arch/mips/cavium-octeon/Makefile                         | 1 -
 drivers/usb/dwc3/Kconfig                                 | 9 +++++++++
 drivers/usb/dwc3/Makefile                                | 1 +
 .../octeon-usb.c => drivers/usb/dwc3/dwc3-octeon.c       | 0
 4 files changed, 10 insertions(+), 1 deletion(-)
 rename arch/mips/cavium-octeon/octeon-usb.c => drivers/usb/dwc3/dwc3-octeon.c (100%)
arch/mips/cavium-octeon/Makefile +0 -1
index 7c02e542959a..2a5926578841 100644
@@ -18,4 +18,3 @@ obj-y += crypto/
 obj-$(CONFIG_MTD)		      += flash_setup.o
 obj-$(CONFIG_SMP)		      += smp.o
 obj-$(CONFIG_OCTEON_ILM)	      += oct_ilm.o
-obj-$(CONFIG_USB)		      += octeon-usb.o
drivers/usb/dwc3/Kconfig +9 -0
index be954a9abbe0..8fc7b7ff7f16 100644
@@ -168,4 +168,13 @@ config USB_DWC3_AM62
 	  The Designware Core USB3 IP is programmed to operate in
 	  in USB 2.0 mode only.
 	  Say 'Y' or 'M' here if you have one such device
+
+config USB_DWC3_OCTEON
+	tristate "Cavium Octeon Platforms"
+	depends on CAVIUM_OCTEON_SOC || COMPILE_TEST
+	default USB_DWC3
+	help
+	  Support Cavium Octeon platforms with DesignWare Core USB3 IP.
+	  Say 'Y' or 'M' here if you have one such device.
+
 endif
drivers/usb/dwc3/Makefile +1 -0
index 9f66bd82b639..fe1493d4bbe5 100644
@@ -54,3 +54,4 @@ obj-$(CONFIG_USB_DWC3_ST)		+= dwc3-st.o
 obj-$(CONFIG_USB_DWC3_QCOM)		+= dwc3-qcom.o
 obj-$(CONFIG_USB_DWC3_IMX8MP)		+= dwc3-imx8mp.o
 obj-$(CONFIG_USB_DWC3_XILINX)		+= dwc3-xilinx.o
+obj-$(CONFIG_USB_DWC3_OCTEON)		+= dwc3-octeon.o
drivers/usb/dwc3/dwc3-octeon.c renamed from arch/mips/cavium-octeon/octeon-usb.c +0 -0
similarity index 100%
rename from arch/mips/cavium-octeon/octeon-usb.c
rename to drivers/usb/dwc3/dwc3-octeon.c
Signature · 1 line
2.39.2
maintainer Acked-by on v1 8/11
#9 lore raw

Re: [PATCH 08/11] usb: dwc3: Move Octeon glue code from arch/mips

Quoted 5 lines · Ladislav Michl

From: Ladislav Michl <[email protected]>

Octeon DWC3 glue code now compiles on all platforms, so move it to drivers/usb/dwc3. No functional changes.

Quoted 8 lines
---
 arch/mips/cavium-octeon/Makefile                         | 1 -
 drivers/usb/dwc3/Kconfig                                 | 9 +++++++++
 drivers/usb/dwc3/Makefile                                | 1 +
 .../octeon-usb.c => drivers/usb/dwc3/dwc3-octeon.c       | 0
 4 files changed, 10 insertions(+), 1 deletion(-)
 rename arch/mips/cavium-octeon/octeon-usb.c => drivers/usb/dwc3/dwc3-octeon.c (100%)
Signature · 2 lines
Crap can work. Given enough thrust pigs will fly, but it's not necessarily a
good idea.                                                [ RFC1925, 2.3 ]
maintainer
#10 lore raw

Re: [PATCH 08/11] usb: dwc3: Move Octeon glue code from arch/mips

In reply to Ladislav Michl
Quoted 29 lines · Ladislav Michl

From: Ladislav Michl <[email protected]>

Octeon DWC3 glue code now compiles on all platforms, so move it to drivers/usb/dwc3. No functional changes.

---
 arch/mips/cavium-octeon/Makefile                         | 1 -
 drivers/usb/dwc3/Kconfig                                 | 9 +++++++++
 drivers/usb/dwc3/Makefile                                | 1 +
 .../octeon-usb.c => drivers/usb/dwc3/dwc3-octeon.c       | 0
 4 files changed, 10 insertions(+), 1 deletion(-)
 rename arch/mips/cavium-octeon/octeon-usb.c => drivers/usb/dwc3/dwc3-octeon.c (100%)
arch/mips/cavium-octeon/Makefile +0 -1
index 7c02e542959a..2a5926578841 100644
@@ -18,4 +18,3 @@ obj-y += crypto/
 obj-$(CONFIG_MTD)		      += flash_setup.o
 obj-$(CONFIG_SMP)		      += smp.o
 obj-$(CONFIG_OCTEON_ILM)	      += oct_ilm.o
-obj-$(CONFIG_USB)		      += octeon-usb.o
drivers/usb/dwc3/Kconfig +0 -0
index be954a9abbe0..8fc7b7ff7f16 100644
@@ -168,4 +168,13 @@ config USB_DWC3_AM62
 	  The Designware Core USB3 IP is programmed to operate in

drivers/usb/dwc3/Kconfig config USB_DWC3_AM62 · line 178

Quoted 10 lines
drivers/usb/dwc3/Kconfig +8 -0
 	  in USB 2.0 mode only.
 	  Say 'Y' or 'M' here if you have one such device
+
+config USB_DWC3_OCTEON
+	tristate "Cavium Octeon Platforms"
+	depends on CAVIUM_OCTEON_SOC || COMPILE_TEST
+	default USB_DWC3
+	help
+	  Support Cavium Octeon platforms with DesignWare Core USB3 IP.
+	  Say 'Y' or 'M' here if you have one such device.

Is this for device, host, or DRD mode? Which IP (DWC_usb3 or DWC_usb31?) and version is it? The more description the better.

Thanks,
Thinh

Quoted 18 lines
drivers/usb/dwc3/Kconfig +1 -0
+
 endif
drivers/usb/dwc3/Makefile +1 -0
index 9f66bd82b639..fe1493d4bbe5 100644
@@ -54,3 +54,4 @@ obj-$(CONFIG_USB_DWC3_ST)		+= dwc3-st.o
 obj-$(CONFIG_USB_DWC3_QCOM)		+= dwc3-qcom.o
 obj-$(CONFIG_USB_DWC3_IMX8MP)		+= dwc3-imx8mp.o
 obj-$(CONFIG_USB_DWC3_XILINX)		+= dwc3-xilinx.o
+obj-$(CONFIG_USB_DWC3_OCTEON)		+= dwc3-octeon.o
drivers/usb/dwc3/dwc3-octeon.c renamed from arch/mips/cavium-octeon/octeon-usb.c +0 -0
similarity index 100%
rename from arch/mips/cavium-octeon/octeon-usb.c
rename to drivers/usb/dwc3/dwc3-octeon.c
Signature · 1 line
2.39.2
#11 lore raw

PATCH 9/11 usb: dwc3: dwc3-octeon: Convert to glue driver

In reply to Ladislav Michl

From: Ladislav Michl <[email protected]>

Use Octeon specific DWC3 glue instead of dwc3-of-simple.

---
 arch/mips/cavium-octeon/octeon-platform.c |   1 -
 drivers/usb/dwc3/dwc3-octeon.c            | 100 +++++++++++-----------
 drivers/usb/dwc3/dwc3-of-simple.c         |   1 -
 3 files changed, 52 insertions(+), 50 deletions(-)
arch/mips/cavium-octeon/octeon-platform.c +0 -1
index ce05c0dd3acd..235c77ce7b18 100644
@@ -450,7 +450,6 @@ static const struct of_device_id octeon_ids[] __initconst = {
 	{ .compatible = "cavium,octeon-3860-bootbus", },
 	{ .compatible = "cavium,mdio-mux", },
 	{ .compatible = "gpio-leds", },
-	{ .compatible = "cavium,octeon-7130-usb-uctl", },
 	{},
 };
 
drivers/usb/dwc3/dwc3-octeon.c +50 -48

115 lines not on this page · raw message

drivers/usb/dwc3/dwc3-of-simple.c +0 -1
index 71fd620c5161..e3423fbea3ed 100644
@@ -172,7 +172,6 @@ static const struct dev_pm_ops dwc3_of_simple_dev_pm_ops = {
 
 static const struct of_device_id of_dwc3_simple_match[] = {
 	{ .compatible = "rockchip,rk3399-dwc3" },
-	{ .compatible = "cavium,octeon-7130-usb-uctl" },
 	{ .compatible = "sprd,sc9860-dwc3" },
 	{ .compatible = "allwinner,sun50i-h6-dwc3" },
 	{ .compatible = "hisilicon,hi3670-dwc3" },
Signature · 1 line
2.39.2
maintainer
#12 lore raw

Re: [PATCH 09/11] usb: dwc3: dwc3-octeon: Convert to glue driver

Quoted 88 lines · Ladislav Michl

From: Ladislav Michl <[email protected]>

Use Octeon specific DWC3 glue instead of dwc3-of-simple.

---
 arch/mips/cavium-octeon/octeon-platform.c |   1 -
 drivers/usb/dwc3/dwc3-octeon.c            | 100 +++++++++++-----------
 drivers/usb/dwc3/dwc3-of-simple.c         |   1 -
 3 files changed, 52 insertions(+), 50 deletions(-)
arch/mips/cavium-octeon/octeon-platform.c +0 -1
index ce05c0dd3acd..235c77ce7b18 100644
@@ -450,7 +450,6 @@ static const struct of_device_id octeon_ids[] __initconst = {
 	{ .compatible = "cavium,octeon-3860-bootbus", },
 	{ .compatible = "cavium,mdio-mux", },
 	{ .compatible = "gpio-leds", },
-	{ .compatible = "cavium,octeon-7130-usb-uctl", },
 	{},
 };
 
drivers/usb/dwc3/dwc3-octeon.c +11 -36
index 2add435ad038..3ebcf2a61233 100644
@@ -187,7 +187,10 @@
 #define USBDRD_UCTL_ECC				0xf0
 #define USBDRD_UCTL_SPARE1			0xf8
 
-static DEFINE_MUTEX(dwc3_octeon_clocks_mutex);
+struct dwc3_data {
+	struct device *dev;
+	void __iomem *base;
+};
 
 #ifdef CONFIG_CAVIUM_OCTEON_SOC
 #include <asm/octeon/octeon.h>
@@ -494,58 +499,57 @@ static void __init dwc3_octeon_phy_reset(void __iomem *base)
 	dwc3_octeon_writeq(uctl_ctl_reg, val);
 }
 
-static int __init dwc3_octeon_device_init(void)
+static int dwc3_octeon_probe(struct platform_device *pdev)
 {
-	const char compat_node_name[] = "cavium,octeon-7130-usb-uctl";
-	struct platform_device *pdev;
-	struct device_node *node;
-	struct resource *res;
-	void __iomem *base;
+	struct device *dev = &pdev->dev;
+	struct dwc3_data *data;
+	int err;
 
-	/*
-	 * There should only be three universal controllers, "uctl"
-	 * in the device tree. Two USB and a SATA, which we ignore.
-	 */
-	node = NULL;
-	do {
-		node = of_find_node_by_name(node, "uctl");
-		if (!node)
-			return -ENODEV;
-
-		if (of_device_is_compatible(node, compat_node_name)) {
-			pdev = of_find_device_by_node(node);
-			if (!pdev)
-				return -ENODEV;
-
-			/*
-			 * The code below maps in the registers necessary for
-			 * setting up the clocks and reseting PHYs. We must
-			 * release the resources so the dwc3 subsystem doesn't
-			 * know the difference.
-			 */
-			base = devm_platform_get_and_ioremap_resource(pdev, 0, &res);
-			if (IS_ERR(base)) {
-				put_device(&pdev->dev);
-				return PTR_ERR(base);
-			}
+	data = devm_kzalloc(dev, sizeof(*data), GFP_KERNEL);
+	if (!data)
+		return -ENOMEM;
 
-			mutex_lock(&dwc3_octeon_clocks_mutex);
-			if (dwc3_octeon_clocks_start(&pdev->dev, base) == 0)
-				dev_info(&pdev->dev, "clocks initialized.\n");

drivers/usb/dwc3/dwc3-octeon.c static void __init dwc3_octeon_phy_reset(void __iomem *base) · line 512

Quoted 10 lines
drivers/usb/dwc3/dwc3-octeon.c +1 -9
-			dwc3_octeon_set_endian_mode(base);
-			dwc3_octeon_phy_reset(base);
-			mutex_unlock(&dwc3_octeon_clocks_mutex);
-			devm_iounmap(&pdev->dev, base);
-			devm_release_mem_region(&pdev->dev, res->start,
-						resource_size(res));
-			put_device(&pdev->dev);
-		}
-	} while (node != NULL);
+	data->dev = dev;

Set private data at the end of this probe function.

Quoted 13 lines
drivers/usb/dwc3/dwc3-octeon.c +10 -1
+	platform_set_drvdata(pdev, data);
 
 
-	return 0;
+	data->base = devm_platform_ioremap_resource(pdev, 0);
+	if (IS_ERR(data->base))
+		return PTR_ERR(data->base);
+
+	err = dwc3_octeon_clocks_start(dev, data->base);
+	if (err)
+		return err;
+
+	dwc3_octeon_set_endian_mode(data->base);

drivers/usb/dwc3/dwc3-octeon.c static void __init dwc3_octeon_phy_reset(void __iomem *base) · line 533

Quoted 10 lines
drivers/usb/dwc3/dwc3-octeon.c +10 -0
+	dwc3_octeon_phy_reset(data->base);
+
+	return of_platform_populate(node, NULL, NULL, dev);
+}
+
+static void dwc3_octeon_remove(struct platform_device *pdev)
+{
+	struct dwc3_data *data = platform_get_drvdata(pdev);
+
+	of_platform_depopulate(data->dev);

Clear private data here:

	platform_set_drvdata(pdev, NULL);
Quoted 39 lines
drivers/usb/dwc3/dwc3-octeon.c +18 -2
 }
-device_initcall(dwc3_octeon_device_init);
 
+static const struct of_device_id dwc3_octeon_of_match[] = {
+	{ .compatible = "cavium,octeon-7130-usb-uctl" },
+	{ },
+};
+MODULE_DEVICE_TABLE(of, dwc3_octeon_of_match);
+
+static struct platform_driver dwc3_octeon_driver = {
+	.probe		= dwc3_octeon_probe,
+	.remove_new	= dwc3_octeon_remove,
+	.driver		= {
+		.name	= "dwc3-octeon",
+		.of_match_table = dwc3_octeon_of_match,
+	},
+};
+module_platform_driver(dwc3_octeon_driver);
+
+MODULE_ALIAS("platform:dwc3-octeon");
 MODULE_AUTHOR("David Daney <[email protected]>");
 MODULE_LICENSE("GPL");
-MODULE_DESCRIPTION("USB driver for OCTEON III SoC");
+MODULE_DESCRIPTION("DesignWare USB3 OCTEON III Glue Layer");
drivers/usb/dwc3/dwc3-of-simple.c +0 -1
index 71fd620c5161..e3423fbea3ed 100644
@@ -172,7 +172,6 @@ static const struct dev_pm_ops dwc3_of_simple_dev_pm_ops = {
 
 static const struct of_device_id of_dwc3_simple_match[] = {
 	{ .compatible = "rockchip,rk3399-dwc3" },
-	{ .compatible = "cavium,octeon-7130-usb-uctl" },
 	{ .compatible = "sprd,sc9860-dwc3" },
 	{ .compatible = "allwinner,sun50i-h6-dwc3" },
 	{ .compatible = "hisilicon,hi3670-dwc3" },
Signature · 1 line
2.39.2

BR,
Thinh

#13 lore raw

PATCH 10/11 usb: dwc3: dwc3-octeon: Move node parsing into driver probe

In reply to Ladislav Michl

From: Ladislav Michl <[email protected]>

Make dwc3_octeon_clocks_start just start the clocks.

---
 drivers/usb/dwc3/dwc3-octeon.c | 150 ++++++++++++++++-----------------
 1 file changed, 71 insertions(+), 79 deletions(-)
drivers/usb/dwc3/dwc3-octeon.c +71 -77
index 3ebcf2a61233..4ad2d8887cf0 100644
@@ -295,67 +295,14 @@ static int dwc3_octeon_config_power(struct device *dev, void __iomem *base)
 	return 0;
 }
 
-static int dwc3_octeon_clocks_start(struct device *dev, void __iomem *base)
+static int dwc3_octeon_clocks_start(struct device *dev, void __iomem *base,
+				    int ref_clk_sel, int ref_clk_fsel,
+				    int mpll_mul)
 {
-	int i, div, mpll_mul, ref_clk_fsel, ref_clk_sel = 2;
-	u32 clock_rate;
+	int div;
 	u64 val;
 	void __iomem *uctl_ctl_reg = base + USBDRD_UCTL_CTL;
 
-	if (dev->of_node) {
-		const char *ss_clock_type;
-		const char *hs_clock_type;
-
-		i = of_property_read_u32(dev->of_node,
-					 "refclk-frequency", &clock_rate);
-		if (i) {
-			dev_err(dev, "No UCTL \"refclk-frequency\"\n");
-			return -EINVAL;
-		}
-		i = of_property_read_string(dev->of_node,
-					    "refclk-type-ss", &ss_clock_type);
-		if (i) {
-			dev_err(dev, "No UCTL \"refclk-type-ss\"\n");
-			return -EINVAL;
-		}
-		i = of_property_read_string(dev->of_node,
-					    "refclk-type-hs", &hs_clock_type);
-		if (i) {
-			dev_err(dev, "No UCTL \"refclk-type-hs\"\n");
-			return -EINVAL;
-		}
-		if (strcmp("dlmc_ref_clk0", ss_clock_type) == 0) {
-			if (strcmp(hs_clock_type, "dlmc_ref_clk0") == 0)
-				ref_clk_sel = 0;
-			else if (strcmp(hs_clock_type, "pll_ref_clk") == 0)
-				ref_clk_sel = 2;
-			else
-				dev_warn(dev, "Invalid HS clock type %s, using pll_ref_clk instead\n",
-					 hs_clock_type);
-		} else if (strcmp(ss_clock_type, "dlmc_ref_clk1") == 0) {
-			if (strcmp(hs_clock_type, "dlmc_ref_clk1") == 0)
-				ref_clk_sel = 1;
-			else if (strcmp(hs_clock_type, "pll_ref_clk") == 0)
-				ref_clk_sel = 3;
-			else {
-				dev_warn(dev, "Invalid HS clock type %s, using pll_ref_clk instead\n",
-					 hs_clock_type);
-				ref_clk_sel = 3;
-			}
-		} else
-			dev_warn(dev, "Invalid SS clock type %s, using dlmc_ref_clk0 instead\n",
-				 ss_clock_type);
-
-		if ((ref_clk_sel == 0 || ref_clk_sel == 1) &&
-		    (clock_rate != 100000000))
-			dev_warn(dev, "Invalid UCTL clock rate of %u, using 100000000 instead\n",
-				 clock_rate);
-
-	} else {
-		dev_err(dev, "No USB UCTL device node\n");
-		return -EINVAL;
-	}
-
 	/*
 	 * Step 1: Wait for all voltages to be stable...that surely
 	 *         happened before starting the kernel. SKIP
@@ -399,24 +346,6 @@ static int dwc3_octeon_clocks_start(struct device *dev, void __iomem *base)
 	val &= ~USBDRD_UCTL_CTL_REF_CLK_SEL;
 	val |= FIELD_PREP(USBDRD_UCTL_CTL_REF_CLK_SEL, ref_clk_sel);
 
-	ref_clk_fsel = 0x07;
-	switch (clock_rate) {
-	default:
-		dev_warn(dev, "Invalid ref_clk %u, using 100000000 instead\n",
-			 clock_rate);
-		fallthrough;
-	case 100000000:
-		mpll_mul = 0x19;
-		if (ref_clk_sel < 2)
-			ref_clk_fsel = 0x27;
-		break;
-	case 50000000:
-		mpll_mul = 0x32;
-		break;
-	case 125000000:
-		mpll_mul = 0x28;
-		break;
-	}
 	val &= ~USBDRD_UCTL_CTL_REF_CLK_FSEL;
 	val |= FIELD_PREP(USBDRD_UCTL_CTL_REF_CLK_FSEL, ref_clk_fsel);
 
@@ -502,8 +429,72 @@ static void __init dwc3_octeon_phy_reset(void __iomem *base)
 static int dwc3_octeon_probe(struct platform_device *pdev)
 {
 	struct device *dev = &pdev->dev;
+	struct device_node *node = dev->of_node;
 	struct dwc3_data *data;
-	int err;
+	int err, ref_clk_sel, ref_clk_fsel, mpll_mul;
+	uint32_t clock_rate;
+	const char *hs_clock_type, *ss_clock_type;
+
+	if (!node) {
+		dev_err(dev, "No USB UCTL device node\n");
+		return -EINVAL;
+	}
+
+	if (of_property_read_u32(node, "refclk-frequency", &clock_rate)) {
+		dev_err(dev, "No UCTL \"refclk-frequency\"\n");
+		return -EINVAL;
+	}
+	if (of_property_read_string(node, "refclk-type-ss", &ss_clock_type)) {
+		dev_err(dev, "No UCTL \"refclk-type-ss\"\n");
+		return -EINVAL;
+	}
+	if (of_property_read_string(node, "refclk-type-hs", &hs_clock_type)) {
+		dev_err(dev, "No UCTL \"refclk-type-hs\"\n");
+		return -EINVAL;
+	}
+
+	ref_clk_sel = 2;
+	if (strcmp("dlmc_ref_clk0", ss_clock_type) == 0) {
+		if (strcmp(hs_clock_type, "dlmc_ref_clk0") == 0)
+			ref_clk_sel = 0;
+		else if (strcmp(hs_clock_type, "pll_ref_clk") == 0)
+			ref_clk_sel = 2;
+		else
+			dev_warn(dev, "Invalid HS clock type %s, using pll_ref_clk instead\n",
+				 hs_clock_type);
+	} else if (strcmp(ss_clock_type, "dlmc_ref_clk1") == 0) {
+		if (strcmp(hs_clock_type, "dlmc_ref_clk1") == 0)
+			ref_clk_sel = 1;
+		else if (strcmp(hs_clock_type, "pll_ref_clk") == 0)
+			ref_clk_sel = 3;
+		else {
+			dev_warn(dev, "Invalid HS clock type %s, using pll_ref_clk instead\n",
+				 hs_clock_type);
+			ref_clk_sel = 3;
+		}
+	} else {
+		dev_warn(dev, "Invalid SS clock type %s, using dlmc_ref_clk0 instead\n",
+			 ss_clock_type);
+	}
+
+	ref_clk_fsel = 0x07;
+	switch (clock_rate) {
+	default:
+		dev_warn(dev, "Invalid ref_clk %u, using 100000000 instead\n",
+			 clock_rate);
+		fallthrough;
+	case 100000000:
+		mpll_mul = 0x19;
+		if (ref_clk_sel < 2)
+			ref_clk_fsel = 0x27;
+		break;
+	case 50000000:
+		mpll_mul = 0x32;
+		break;
+	case 125000000:
+		mpll_mul = 0x28;
+		break;
+	}
 
 	data = devm_kzalloc(dev, sizeof(*data), GFP_KERNEL);
 	if (!data)
@@ -516,7 +507,8 @@ static int dwc3_octeon_probe(struct platform_device *pdev)
 	if (IS_ERR(data->base))
 		return PTR_ERR(data->base);
 
-	err = dwc3_octeon_clocks_start(dev, data->base);
+	err = dwc3_octeon_clocks_start(dev, data->base,
+				       ref_clk_sel, ref_clk_fsel, mpll_mul);
 	if (err)
 		return err;
 
Signature · 1 line
2.39.2
maintainer
#14 lore raw

Re: [PATCH 10/11] usb: dwc3: dwc3-octeon: Move node parsing into driver probe

Quoted 130 lines · Ladislav Michl

From: Ladislav Michl <[email protected]>

Make dwc3_octeon_clocks_start just start the clocks.

---
 drivers/usb/dwc3/dwc3-octeon.c | 150 ++++++++++++++++-----------------
 1 file changed, 71 insertions(+), 79 deletions(-)
drivers/usb/dwc3/dwc3-octeon.c +18 -76
index 3ebcf2a61233..4ad2d8887cf0 100644
@@ -295,67 +295,14 @@ static int dwc3_octeon_config_power(struct device *dev, void __iomem *base)
 	return 0;
 }
 
-static int dwc3_octeon_clocks_start(struct device *dev, void __iomem *base)
+static int dwc3_octeon_clocks_start(struct device *dev, void __iomem *base,
+				    int ref_clk_sel, int ref_clk_fsel,
+				    int mpll_mul)
 {
-	int i, div, mpll_mul, ref_clk_fsel, ref_clk_sel = 2;
-	u32 clock_rate;
+	int div;
 	u64 val;
 	void __iomem *uctl_ctl_reg = base + USBDRD_UCTL_CTL;
 
-	if (dev->of_node) {
-		const char *ss_clock_type;
-		const char *hs_clock_type;
-
-		i = of_property_read_u32(dev->of_node,
-					 "refclk-frequency", &clock_rate);
-		if (i) {
-			dev_err(dev, "No UCTL \"refclk-frequency\"\n");
-			return -EINVAL;
-		}
-		i = of_property_read_string(dev->of_node,
-					    "refclk-type-ss", &ss_clock_type);
-		if (i) {
-			dev_err(dev, "No UCTL \"refclk-type-ss\"\n");
-			return -EINVAL;
-		}
-		i = of_property_read_string(dev->of_node,
-					    "refclk-type-hs", &hs_clock_type);
-		if (i) {
-			dev_err(dev, "No UCTL \"refclk-type-hs\"\n");
-			return -EINVAL;
-		}
-		if (strcmp("dlmc_ref_clk0", ss_clock_type) == 0) {
-			if (strcmp(hs_clock_type, "dlmc_ref_clk0") == 0)
-				ref_clk_sel = 0;
-			else if (strcmp(hs_clock_type, "pll_ref_clk") == 0)
-				ref_clk_sel = 2;
-			else
-				dev_warn(dev, "Invalid HS clock type %s, using pll_ref_clk instead\n",
-					 hs_clock_type);
-		} else if (strcmp(ss_clock_type, "dlmc_ref_clk1") == 0) {
-			if (strcmp(hs_clock_type, "dlmc_ref_clk1") == 0)
-				ref_clk_sel = 1;
-			else if (strcmp(hs_clock_type, "pll_ref_clk") == 0)
-				ref_clk_sel = 3;
-			else {
-				dev_warn(dev, "Invalid HS clock type %s, using pll_ref_clk instead\n",
-					 hs_clock_type);
-				ref_clk_sel = 3;
-			}
-		} else
-			dev_warn(dev, "Invalid SS clock type %s, using dlmc_ref_clk0 instead\n",
-				 ss_clock_type);
-
-		if ((ref_clk_sel == 0 || ref_clk_sel == 1) &&
-		    (clock_rate != 100000000))
-			dev_warn(dev, "Invalid UCTL clock rate of %u, using 100000000 instead\n",
-				 clock_rate);
-
-	} else {
-		dev_err(dev, "No USB UCTL device node\n");
-		return -EINVAL;
-	}
-
 	/*
 	 * Step 1: Wait for all voltages to be stable...that surely
 	 *         happened before starting the kernel. SKIP
@@ -399,24 +346,6 @@ static int dwc3_octeon_clocks_start(struct device *dev, void __iomem *base)
 	val &= ~USBDRD_UCTL_CTL_REF_CLK_SEL;
 	val |= FIELD_PREP(USBDRD_UCTL_CTL_REF_CLK_SEL, ref_clk_sel);
 
-	ref_clk_fsel = 0x07;
-	switch (clock_rate) {
-	default:
-		dev_warn(dev, "Invalid ref_clk %u, using 100000000 instead\n",
-			 clock_rate);
-		fallthrough;
-	case 100000000:
-		mpll_mul = 0x19;
-		if (ref_clk_sel < 2)
-			ref_clk_fsel = 0x27;
-		break;
-	case 50000000:
-		mpll_mul = 0x32;
-		break;
-	case 125000000:
-		mpll_mul = 0x28;
-		break;
-	}
 	val &= ~USBDRD_UCTL_CTL_REF_CLK_FSEL;
 	val |= FIELD_PREP(USBDRD_UCTL_CTL_REF_CLK_FSEL, ref_clk_fsel);
 
@@ -502,8 +429,72 @@ static void __init dwc3_octeon_phy_reset(void __iomem *base)
 static int dwc3_octeon_probe(struct platform_device *pdev)
 {
 	struct device *dev = &pdev->dev;
+	struct device_node *node = dev->of_node;
 	struct dwc3_data *data;
-	int err;
+	int err, ref_clk_sel, ref_clk_fsel, mpll_mul;
+	uint32_t clock_rate;
+	const char *hs_clock_type, *ss_clock_type;
+
+	if (!node) {
+		dev_err(dev, "No USB UCTL device node\n");
+		return -EINVAL;
+	}
+
+	if (of_property_read_u32(node, "refclk-frequency", &clock_rate)) {
+		dev_err(dev, "No UCTL \"refclk-frequency\"\n");
+		return -EINVAL;
+	}

drivers/usb/dwc3/dwc3-octeon.c static void __init dwc3_octeon_phy_reset(void __iomem *base) · line 456

Quoted 10 lines
drivers/usb/dwc3/dwc3-octeon.c +10 -0
+	if (of_property_read_string(node, "refclk-type-ss", &ss_clock_type)) {
+		dev_err(dev, "No UCTL \"refclk-type-ss\"\n");
+		return -EINVAL;
+	}
+	if (of_property_read_string(node, "refclk-type-hs", &hs_clock_type)) {
+		dev_err(dev, "No UCTL \"refclk-type-hs\"\n");
+		return -EINVAL;
+	}
+
+	ref_clk_sel = 2;

Can we use macro instead of magic numbers?

drivers/usb/dwc3/dwc3-octeon.c static void __init dwc3_octeon_phy_reset(void __iomem *base) · line 469

Quoted 13 lines
drivers/usb/dwc3/dwc3-octeon.c +13 -0
+	if (strcmp("dlmc_ref_clk0", ss_clock_type) == 0) {
+		if (strcmp(hs_clock_type, "dlmc_ref_clk0") == 0)
+			ref_clk_sel = 0;
+		else if (strcmp(hs_clock_type, "pll_ref_clk") == 0)
+			ref_clk_sel = 2;
+		else
+			dev_warn(dev, "Invalid HS clock type %s, using pll_ref_clk instead\n",
+				 hs_clock_type);
+	} else if (strcmp(ss_clock_type, "dlmc_ref_clk1") == 0) {
+		if (strcmp(hs_clock_type, "dlmc_ref_clk1") == 0)
+			ref_clk_sel = 1;
+		else if (strcmp(hs_clock_type, "pll_ref_clk") == 0)
+			ref_clk_sel = 3;

Please run checkpatch and fix minor formatting issues such as bracket here.

Quoted 44 lines
drivers/usb/dwc3/dwc3-octeon.c +30 -1
+		else {
+			dev_warn(dev, "Invalid HS clock type %s, using pll_ref_clk instead\n",
+				 hs_clock_type);
+			ref_clk_sel = 3;
+		}
+	} else {
+		dev_warn(dev, "Invalid SS clock type %s, using dlmc_ref_clk0 instead\n",
+			 ss_clock_type);
+	}
+
+	ref_clk_fsel = 0x07;
+	switch (clock_rate) {
+	default:
+		dev_warn(dev, "Invalid ref_clk %u, using 100000000 instead\n",
+			 clock_rate);
+		fallthrough;
+	case 100000000:
+		mpll_mul = 0x19;
+		if (ref_clk_sel < 2)
+			ref_clk_fsel = 0x27;
+		break;
+	case 50000000:
+		mpll_mul = 0x32;
+		break;
+	case 125000000:
+		mpll_mul = 0x28;
+		break;
+	}
 
 	data = devm_kzalloc(dev, sizeof(*data), GFP_KERNEL);
 	if (!data)
@@ -516,7 +507,8 @@ static int dwc3_octeon_probe(struct platform_device *pdev)
 	if (IS_ERR(data->base))
 		return PTR_ERR(data->base);
 
-	err = dwc3_octeon_clocks_start(dev, data->base);
+	err = dwc3_octeon_clocks_start(dev, data->base,
+				       ref_clk_sel, ref_clk_fsel, mpll_mul);
 	if (err)
 		return err;
 
Signature · 1 line
2.39.2

Thanks,
Thinh

#15 lore raw

Re: [PATCH 10/11] usb: dwc3: dwc3-octeon: Move node parsing into driver probe

Hi,

thank you for review, v2 is about to be sent, here just let me anwer issue I didn't address there.

Quoted 1 line · Thinh Nguyen

On Mon, Jun 19, 2023, Ladislav Michl wrote:

[snip]

Quoted 3 lines

Review

Quoted 1 line
(quoted diff) +1 -0
+	ref_clk_sel = 2;

Can we use macro instead of magic numbers?

This is a bit problematic, comment above USBDRD_UCTL_CTL_REF_CLK_SEL says:
/* Reference clock select for SuperSpeed and HighSpeed PLLs:

 *      0x0 = Both PLLs use DLMC_REF_CLK0 for reference clock
 *      0x1 = Both PLLs use DLMC_REF_CLK1 for reference clock
 *      0x2 = SuperSpeed PLL uses DLMC_REF_CLK0 for reference clock &

* HighSpeed PLL uses PLL_REF_CLK for reference clck
* 0x3 = SuperSpeed PLL uses DLMC_REF_CLK1 for reference clock &
* HighSpeed PLL uses PLL_REF_CLK for reference clck
*/
So I really cannot imagine sane name. Also please note, that field and register names were created to match documentation. Values, however, have no defined names, so I would need to invent them. That is something I'd like to avoid as it might be confusing.

And last, but to least: Octeon would deserve proper clock api (clk driver to be implemented). So hopefully that code get replaced with more sane one. For now, I just kept compatibility with current DT.

Thanks,
ladis

maintainer
#16 lore raw

Re: [PATCH 10/11] usb: dwc3: dwc3-octeon: Move node parsing into driver probe

Quoted 27 lines · Ladislav Michl

Hi,

thank you for review, v2 is about to be sent, here just let me anwer issue I didn't address there.

Quoted 1 line · Thinh Nguyen

On Mon, Jun 19, 2023, Ladislav Michl wrote:

[snip]

Quoted 3 lines

Review

Quoted 1 line
(quoted diff) +1 -0
+	ref_clk_sel = 2;

Can we use macro instead of magic numbers?

This is a bit problematic, comment above USBDRD_UCTL_CTL_REF_CLK_SEL says:
/* Reference clock select for SuperSpeed and HighSpeed PLLs:

 *      0x0 = Both PLLs use DLMC_REF_CLK0 for reference clock
 *      0x1 = Both PLLs use DLMC_REF_CLK1 for reference clock
 *      0x2 = SuperSpeed PLL uses DLMC_REF_CLK0 for reference clock &

* HighSpeed PLL uses PLL_REF_CLK for reference clck
* 0x3 = SuperSpeed PLL uses DLMC_REF_CLK1 for reference clock &
* HighSpeed PLL uses PLL_REF_CLK for reference clck
*/
So I really cannot imagine sane name. Also please note, that field and register names were created to match documentation. Values, however, have no defined names, so I would need to invent them. That is something I'd like to avoid as it might be confusing.

Quoted 7 lines

And last, but to least: Octeon would deserve proper clock api (clk driver to be implemented). So hopefully that code get replaced with more sane one. For now, I just kept compatibility with current DT.

Thanks,
ladis

Sure. That's reasonable.

Thanks,
Thinh

#17 lore raw

Re: [PATCH 10/11] usb: dwc3: dwc3-octeon: Move node parsing into driver probe

In reply to Ladislav Michl
kernel test robot noticed the following build warnings:

Hi Ladislav,

kernel test robot noticed the following build warnings:

[auto build test WARNING on usb/usb-testing]
[also build test WARNING on usb/usb-next usb/usb-linus westeri-thunderbolt/next v6.4] [cannot apply to linus/master next-20230630]
[If your patch is applied to the wrong git tree, kindly drop us a note. And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch#_base_tree_information]

url:    https://github.com/intel-lab-lkp/linux/commits/Ladislav-Michl/MIPS-OCTEON-octeon-usb-add-all-register-offsets/20230620-041822
base:   https://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb.git usb-testing
patch link:    https://lore.kernel.org/r/ZJC3eK8QMxShyZDt%40lenoch
patch subject: [PATCH 10/11] usb: dwc3: dwc3-octeon: Move node parsing into driver probe
config: mips-randconfig-r083-20230701 (https://download.01.org/0day-ci/archive/20230701/[email protected]/config)
compiler: mips64-linux-gcc (GCC) 12.3.0
reproduce: (https://download.01.org/0day-ci/archive/20230701/[email protected]/reproduce)

If you fix the issue in a separate patch/commit (i.e. not just a new version of the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <[email protected]>
| Closes: https://lore.kernel.org/oe-kbuild-all/[email protected]/

sparse warnings: (new ones prefixed by >>)

>> drivers/usb/dwc3/dwc3-octeon.c:277:43: sparse: sparse: cast removes address space '__iomem' of expression

vim +/__iomem +277 drivers/usb/dwc3/dwc3-octeon.c

d83bf20c53410d arch/mips/cavium-octeon/octeon-usb.c Ladislav Michl 2023-06-19  255  
3c47bbb8f554f8 arch/mips/cavium-octeon/octeon-usb.c Ladislav Michl 2023-06-19  256  static int dwc3_octeon_config_power(struct device *dev, void __iomem *base)
06df6469e3e1a1 arch/mips/cavium-octeon/octeon-usb.c Ladislav Michl 2023-06-19  257  {
06df6469e3e1a1 arch/mips/cavium-octeon/octeon-usb.c Ladislav Michl 2023-06-19  258  	uint32_t gpio_pwr[3];
06df6469e3e1a1 arch/mips/cavium-octeon/octeon-usb.c Ladislav Michl 2023-06-19  259  	int gpio, len, power_active_low;
06df6469e3e1a1 arch/mips/cavium-octeon/octeon-usb.c Ladislav Michl 2023-06-19  260  	struct device_node *node = dev->of_node;
06df6469e3e1a1 arch/mips/cavium-octeon/octeon-usb.c Ladislav Michl 2023-06-19  261  	u64 val;
3c47bbb8f554f8 arch/mips/cavium-octeon/octeon-usb.c Ladislav Michl 2023-06-19  262  	void __iomem *uctl_host_cfg_reg = base + USBDRD_UCTL_HOST_CFG;
06df6469e3e1a1 arch/mips/cavium-octeon/octeon-usb.c Ladislav Michl 2023-06-19  263  
06df6469e3e1a1 arch/mips/cavium-octeon/octeon-usb.c Ladislav Michl 2023-06-19  264  	if (of_find_property(node, "power", &len) != NULL) {
06df6469e3e1a1 arch/mips/cavium-octeon/octeon-usb.c Ladislav Michl 2023-06-19  265  		if (len == 12) {
06df6469e3e1a1 arch/mips/cavium-octeon/octeon-usb.c Ladislav Michl 2023-06-19  266  			of_property_read_u32_array(node, "power", gpio_pwr, 3);
06df6469e3e1a1 arch/mips/cavium-octeon/octeon-usb.c Ladislav Michl 2023-06-19  267  			power_active_low = gpio_pwr[2] & 0x01;
06df6469e3e1a1 arch/mips/cavium-octeon/octeon-usb.c Ladislav Michl 2023-06-19  268  			gpio = gpio_pwr[1];
06df6469e3e1a1 arch/mips/cavium-octeon/octeon-usb.c Ladislav Michl 2023-06-19  269  		} else if (len == 8) {
06df6469e3e1a1 arch/mips/cavium-octeon/octeon-usb.c Ladislav Michl 2023-06-19  270  			of_property_read_u32_array(node, "power", gpio_pwr, 2);
06df6469e3e1a1 arch/mips/cavium-octeon/octeon-usb.c Ladislav Michl 2023-06-19  271  			power_active_low = 0;
06df6469e3e1a1 arch/mips/cavium-octeon/octeon-usb.c Ladislav Michl 2023-06-19  272  			gpio = gpio_pwr[1];
06df6469e3e1a1 arch/mips/cavium-octeon/octeon-usb.c Ladislav Michl 2023-06-19  273  		} else {
06df6469e3e1a1 arch/mips/cavium-octeon/octeon-usb.c Ladislav Michl 2023-06-19  274  			dev_err(dev, "invalid power configuration\n");
06df6469e3e1a1 arch/mips/cavium-octeon/octeon-usb.c Ladislav Michl 2023-06-19  275  			return -EINVAL;
06df6469e3e1a1 arch/mips/cavium-octeon/octeon-usb.c Ladislav Michl 2023-06-19  276  		}
3c47bbb8f554f8 arch/mips/cavium-octeon/octeon-usb.c Ladislav Michl 2023-06-19 @277  		dwc3_octeon_config_gpio(((u64)base >> 24) & 1, gpio);
93e502b3c2d44d arch/mips/cavium-octeon/octeon-usb.c Steven J. Hill 2017-01-25  278  
93e502b3c2d44d arch/mips/cavium-octeon/octeon-usb.c Steven J. Hill 2017-01-25  279  		/* Enable XHCI power control and set if active high or low. */
3c47bbb8f554f8 arch/mips/cavium-octeon/octeon-usb.c Ladislav Michl 2023-06-19  280  		val = dwc3_octeon_readq(uctl_host_cfg_reg);
52245e391fcf6c arch/mips/cavium-octeon/octeon-usb.c Ladislav Michl 2023-06-19  281  		val |= USBDRD_UCTL_HOST_PPC_EN;
52245e391fcf6c arch/mips/cavium-octeon/octeon-usb.c Ladislav Michl 2023-06-19  282  		if (power_active_low)
52245e391fcf6c arch/mips/cavium-octeon/octeon-usb.c Ladislav Michl 2023-06-19  283  			val &= ~USBDRD_UCTL_HOST_PPC_ACTIVE_HIGH_EN;
52245e391fcf6c arch/mips/cavium-octeon/octeon-usb.c Ladislav Michl 2023-06-19  284  		else
52245e391fcf6c arch/mips/cavium-octeon/octeon-usb.c Ladislav Michl 2023-06-19  285  			val |= USBDRD_UCTL_HOST_PPC_ACTIVE_HIGH_EN;
3c47bbb8f554f8 arch/mips/cavium-octeon/octeon-usb.c Ladislav Michl 2023-06-19  286  		dwc3_octeon_writeq(uctl_host_cfg_reg, val);
93e502b3c2d44d arch/mips/cavium-octeon/octeon-usb.c Steven J. Hill 2017-01-25  287  	} else {
93e502b3c2d44d arch/mips/cavium-octeon/octeon-usb.c Steven J. Hill 2017-01-25  288  		/* Disable XHCI power control and set if active high. */
3c47bbb8f554f8 arch/mips/cavium-octeon/octeon-usb.c Ladislav Michl 2023-06-19  289  		val = dwc3_octeon_readq(uctl_host_cfg_reg);
52245e391fcf6c arch/mips/cavium-octeon/octeon-usb.c Ladislav Michl 2023-06-19  290  		val &= ~USBDRD_UCTL_HOST_PPC_EN;
52245e391fcf6c arch/mips/cavium-octeon/octeon-usb.c Ladislav Michl 2023-06-19  291  		val &= ~USBDRD_UCTL_HOST_PPC_ACTIVE_HIGH_EN;
3c47bbb8f554f8 arch/mips/cavium-octeon/octeon-usb.c Ladislav Michl 2023-06-19  292  		dwc3_octeon_writeq(uctl_host_cfg_reg, val);
4a24f6e0cc17ba arch/mips/cavium-octeon/octeon-usb.c Ladislav Michl 2022-12-21  293  		dev_info(dev, "power control disabled\n");
93e502b3c2d44d arch/mips/cavium-octeon/octeon-usb.c Steven J. Hill 2017-01-25  294  	}
93e502b3c2d44d arch/mips/cavium-octeon/octeon-usb.c Steven J. Hill 2017-01-25  295  	return 0;
93e502b3c2d44d arch/mips/cavium-octeon/octeon-usb.c Steven J. Hill 2017-01-25  296  }
93e502b3c2d44d arch/mips/cavium-octeon/octeon-usb.c Steven J. Hill 2017-01-25  297  
Signature · 2 lines
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki
#18 lore raw

PATCH 11/11 usb: dwc3: Add SPDX header and copyright

In reply to Ladislav Michl

From: Ladislav Michl <[email protected]>

As driver is rewritten and David no longer works for Marvell (Cavium), I'm to blame for breakage.

---
 drivers/usb/dwc3/dwc3-octeon.c | 10 ++++------
 1 file changed, 4 insertions(+), 6 deletions(-)
drivers/usb/dwc3/dwc3-octeon.c +4 -6
index 4ad2d8887cf0..1776bbaf28c0 100644
@@ -1,11 +1,9 @@
+// SPDX-License-Identifier: GPL-2.0
 /*
- * XHCI HCD glue for Cavium Octeon III SOCs.
+ * DWC3 glue for Cavium Octeon III SOCs.
  *
  * Copyright (C) 2010-2017 Cavium Networks
- *
- * This file is subject to the terms and conditions of the GNU General Public
- * License.  See the file "COPYING" in the main directory of this archive
- * for more details.
+ * Copyright (C) 2023 Ladislav Michl <[email protected]>
  */
 
 #include <linux/bitfield.h>
@@ -542,6 +540,6 @@ static struct platform_driver dwc3_octeon_driver = {
 module_platform_driver(dwc3_octeon_driver);
 
 MODULE_ALIAS("platform:dwc3-octeon");
-MODULE_AUTHOR("David Daney <[email protected]>");
+MODULE_AUTHOR("Ladislav Michl <[email protected]>");
 MODULE_LICENSE("GPL");
 MODULE_DESCRIPTION("DesignWare USB3 OCTEON III Glue Layer");
Signature · 1 line
2.39.2
maintainer
#19 lore raw
In reply to Ladislav Michl

Hi,

Quoted 53 lines · Ladislav Michl

Hi,

primary motivation was to fix issue Id 29206 as described in OCTEON III CN70XX/CN71XX Known Issues, version: 1.9. Said document is marked as Marvell Proprietary and Confidential, therefore I'm not sure if I can cite from it.

This probably does not matter too much as the root of the information listed there is a workaround being implemented in OCTEON SDK 3.1.2 patch 7 and later in u-boot/drivers/usb/host/xhci-octeon.c:dwc3_uphy_pll_reset()

My coleague ported that patch to linux-4.9 and it is appended to this cover letter for a reference.

The glue code currently lives in arch/mips/cavium-octeon/octeon-usb.c and loops for each "cavium,octeon-7130-usb-uctl" compatible. However there is no bond with dwc3 core code, so if anything goes wrong in glue code, the loop breaks, leaving dwc3 in reset.

Later on when dwc3 core tries to read any device register, bus error is emited, leading to kernel panic.

Therefore glue code is cleaned-up first, then moved to drivers/usb/dwc3 and modified to match current state of art.

Now it is easier to properly implement said errata. Does it belong to core code as a quirk? Comments and suggestions welcome.

Ladislav Michl (11):
MIPS: OCTEON: octeon-usb: add all register offsets
MIPS: OCTEON: octeon-usb: use bitfields for control register
MIPS: OCTEON: octeon-usb: use bitfields for host config register
MIPS: OCTEON: octeon-usb: use bitfields for shim register
MIPS: OCTEON: octeon-usb: move gpio config to separate function
MIPS: OCTEON: octeon-usb: introduce dwc3_octeon_{read,write}q
MIPS: OCTEON: octeon-usb: cleanup divider calculation
usb: dwc3: Move Octeon glue code from arch/mips
usb: dwc3: dwc3-octeon: Convert to glue driver
usb: dwc3: dwc3-octeon: Move node parsing into driver probe
usb: dwc3: Add SPDX header and copyright

 arch/mips/cavium-octeon/Makefile          |   1 -
 arch/mips/cavium-octeon/octeon-platform.c |   1 -
 arch/mips/cavium-octeon/octeon-usb.c      | 548 ----------------------
 drivers/usb/dwc3/Kconfig                  |   9 +
 drivers/usb/dwc3/Makefile                 |   1 +
 drivers/usb/dwc3/dwc3-octeon.c            | 545 +++++++++++++++++++++
 drivers/usb/dwc3/dwc3-of-simple.c         |   1 -
 7 files changed, 555 insertions(+), 551 deletions(-)
 delete mode 100644 arch/mips/cavium-octeon/octeon-usb.c
 create mode 100644 drivers/usb/dwc3/dwc3-octeon.c
Quoted 1 line

---

Please use get_maintainer.pl to get all the proper maintainers to review the changes.

Since this is a large change involving different subsystems, if possible, please split the changes related to MIPS and try to upstream those first as they will affect how dwc3 glue driver will look.

Thanks,
Thinh

#20 lore raw

Hi,

Quoted 1 line · Thinh Nguyen

Hi,

[snip]

Quoted 2 lines

Please use get_maintainer.pl to get all the proper maintainers to review the changes.

That's what I did. Thomas, MIPS maintainer is reading linux-mips list, Greg is reading linux-usb list, you were the one who receives changes, hence all patcheset was sent to you in a hope you provide some comments about actual code changes.

Quoted 3 lines

Since this is a large change involving different subsystems, if possible, please split the changes related to MIPS and try to upstream those first as they will affect how dwc3 glue driver will look.

That's pretty straightforward as patchset is arranged exactly this way, so MIPS maintainer is free to apply patches till driver move. However, any actual feedback would be still usefull. In case it gets some acks I'll add them and rebase patches to the latest -next.

Quoted 2 lines

Thanks,
Thinh

Thanks,
ladis

#21 lore raw
Quoted 2 lines · Ladislav Michl

That's pretty straightforward as patchset is arranged exactly this way, so MIPS maintainer is free to apply patches till driver move.

just to get things moving I've applied patch 1-7 to mips-next. If I should take the whole thing I need ACKs for it, otherwise I'm going to add an Acked-by for the moving patch (patch 8), which then could be accepted after next merge window.

Thomas.

Signature · 2 lines
Crap can work. Given enough thrust pigs will fly, but it's not necessarily a
good idea.                                                [ RFC1925, 2.3 ]
maintainer
#22 lore raw
In reply to Ladislav Michl
Quoted 4 lines · Ladislav Michl

Hi,

Quoted 1 line · Thinh Nguyen

Hi,

Quoted 8 lines

[snip]

Quoted 2 lines

Please use get_maintainer.pl to get all the proper maintainers to review the changes.

That's what I did. Thomas, MIPS maintainer is reading linux-mips list, Greg is reading linux-usb list, you were the one who receives changes, hence all patcheset was sent to you in a hope you provide some comments about actual code changes.

I can take a look at the changes. However, the patches were sent directly To: me for patches related MIPS and no one else except the linux-usb linux-mips (without Thomas), which doesn't look right.

I'm not familiar with the MIPS subsystem, but typically for the USB mailing list, we capture all the emails generated from get_maintainer.pl and include them To: and Cc: should they need attention to the changes. Often we filter our emails based on whether it's directed toward us, at least I do that. Also, we can't be sure if everyone still subscribes to the corresponding mailing list.

Quoted 10 lines
Quoted 4 lines

Since this is a large change involving different subsystems, if possible, please split the changes related to MIPS and try to upstream those first as they will affect how dwc3 glue driver will look.

That's pretty straightforward as patchset is arranged exactly this way, so MIPS maintainer is free to apply patches till driver move. However, any actual feedback would be still usefull. In case it gets some acks I'll add them and rebase patches to the latest -next.

Then that's good. If the MIPS maintainer approves all the MIPS related changes and already pick them up, there should be no problem.

Thanks,
Thinh

#23 lore raw

Hi,

Quoted 18 lines · Thinh Nguyen
Quoted 4 lines · Ladislav Michl

Hi,

Quoted 1 line · Thinh Nguyen

Hi,

Quoted 8 lines

[snip]

Quoted 2 lines

Please use get_maintainer.pl to get all the proper maintainers to review the changes.

That's what I did. Thomas, MIPS maintainer is reading linux-mips list, Greg is reading linux-usb list, you were the one who receives changes, hence all patcheset was sent to you in a hope you provide some comments about actual code changes.

I can take a look at the changes. However, the patches were sent directly To: me for patches related MIPS and no one else except the linux-usb linux-mips (without Thomas), which doesn't look right.

Quoted 6 lines

I'm not familiar with the MIPS subsystem, but typically for the USB mailing list, we capture all the emails generated from get_maintainer.pl and include them To: and Cc: should they need attention to the changes. Often we filter our emails based on whether it's directed toward us, at least I do that. Also, we can't be sure if everyone still subscribes to the corresponding mailing list.

Did you run get_maintainer.pl yourself? Because here (for v6.3) I get:

$ ./scripts/get_maintainer.pl -f arch/mips/cavium-octeon/octeon-usb.c 
Thomas Bogendoerfer <[email protected]> (maintainer:MIPS,commit_signer:1/1=100%)
Ladislav Michl <[email protected]> (commit_signer:1/1=100%,authored:1/1=100%,added_lines:20/20=100%,removed_lines:22/22=100%)
[email protected] (open list:MIPS)
[email protected] (open list)
$ ./scripts/get_maintainer.pl -f drivers/usb/dwc3
Thinh Nguyen <[email protected]> (maintainer:DESIGNWARE USB3 DRD IP DRIVER)
Greg Kroah-Hartman <[email protected]> (supporter:USB SUBSYSTEM)
[email protected] (open list:DESIGNWARE USB3 DRD IP DRIVER)
[email protected] (open list)

So I have no clue who else should I add to Cc list...

Quoted 5 lines
Quoted 5 lines
Quoted 4 lines

Since this is a large change involving different subsystems, if possible, please split the changes related to MIPS and try to upstream those first as they will affect how dwc3 glue driver will look.

Quoted 8 lines
Quoted 5 lines

That's pretty straightforward as patchset is arranged exactly this way, so MIPS maintainer is free to apply patches till driver move. However, any actual feedback would be still usefull. In case it gets some acks I'll add them and rebase patches to the latest -next.

Then that's good. If the MIPS maintainer approves all the MIPS related changes and already pick them up, there should be no problem.

Well, remaining question is how to make it comfortable for you to take changes. Patches 1-7 are in mips-next already, so I propose to wait until they propagate upstream then I rebase remaining patches to whatever is in linux-next.

Please note I added glue driver into Makefile and Kconfig at random locations as I failed to find any logic here. If you have any preference or want to sort those files alphabetically first for example, just let me know and I'll add some cleanup patches before sending v2.

Also coleagues of mine meanwhile found that PLL indeed ocassionally fails to lock, so workaround attached to cover letter is really needed. Naturally it cannot sneak in as it is, so unless you have better idea I'll just port it to recent driver state and we can start discussion from there in a separate thread.

Thank you,
ladis

maintainer
#24 lore raw

Hi,

Quoted 40 lines · Ladislav Michl

Hi,

Quoted 18 lines · Thinh Nguyen
Quoted 4 lines · Ladislav Michl

Hi,

Quoted 1 line · Thinh Nguyen

Hi,

Quoted 8 lines

[snip]

Quoted 2 lines

Please use get_maintainer.pl to get all the proper maintainers to review the changes.

That's what I did. Thomas, MIPS maintainer is reading linux-mips list, Greg is reading linux-usb list, you were the one who receives changes, hence all patcheset was sent to you in a hope you provide some comments about actual code changes.

I can take a look at the changes. However, the patches were sent directly To: me for patches related MIPS and no one else except the linux-usb linux-mips (without Thomas), which doesn't look right.

Quoted 6 lines

I'm not familiar with the MIPS subsystem, but typically for the USB mailing list, we capture all the emails generated from get_maintainer.pl and include them To: and Cc: should they need attention to the changes. Often we filter our emails based on whether it's directed toward us, at least I do that. Also, we can't be sure if everyone still subscribes to the corresponding mailing list.

Did you run get_maintainer.pl yourself? Because here (for v6.3) I get:

$ ./scripts/get_maintainer.pl -f arch/mips/cavium-octeon/octeon-usb.c 
Thomas Bogendoerfer <[email protected]> (maintainer:MIPS,commit_signer:1/1=100%)
Ladislav Michl <[email protected]> (commit_signer:1/1=100%,authored:1/1=100%,added_lines:20/20=100%,removed_lines:22/22=100%)
[email protected] (open list:MIPS)
[email protected] (open list)
$ ./scripts/get_maintainer.pl -f drivers/usb/dwc3
Thinh Nguyen <[email protected]> (maintainer:DESIGNWARE USB3 DRD IP DRIVER)
Greg Kroah-Hartman <[email protected]> (supporter:USB SUBSYSTEM)
[email protected] (open list:DESIGNWARE USB3 DRD IP DRIVER)
[email protected] (open list)
Quoted 1 line

So I have no clue who else should I add to Cc list...

For each patch, please include all the emails returned from get_maintainer.pl including Greg's and Thomas's.

Quoted 12 lines
Quoted 12 lines
Quoted 11 lines
Quoted 5 lines

Since this is a large change involving different subsystems, if possible, please split the changes related to MIPS and try to upstream those first as they will affect how dwc3 glue driver will look.

That's pretty straightforward as patchset is arranged exactly this way, so MIPS maintainer is free to apply patches till driver move. However, any actual feedback would be still usefull. In case it gets some acks I'll add them and rebase patches to the latest -next.

Quoted 7 lines
Quoted 2 lines

Then that's good. If the MIPS maintainer approves all the MIPS related changes and already pick them up, there should be no problem.

Well, remaining question is how to make it comfortable for you to take changes. Patches 1-7 are in mips-next already, so I propose to wait until they propagate upstream then I rebase remaining patches to whatever is in linux-next.

Yes. Please rebase the remaining patches and resend them.

Quoted 5 lines

Please note I added glue driver into Makefile and Kconfig at random locations as I failed to find any logic here. If you have any preference or want to sort those files alphabetically first for example, just let me know and I'll add some cleanup patches before sending v2.

Unfortunately there isn't really an order to them. What you did is fine.

Quoted 6 lines

Also coleagues of mine meanwhile found that PLL indeed ocassionally fails to lock, so workaround attached to cover letter is really needed. Naturally it cannot sneak in as it is, so unless you have better idea I'll just port it to recent driver state and we can start discussion from there in a separate thread.

If this causes a regression, then please fix it before sending it in. If it's a new found issue, you can create a fix patch at a later point.

Thanks,
Thinh

#25 lore raw

Hi,

Quoted 1 line · Thinh Nguyen

Hi,

[snip]

Quoted 2 lines

For each patch, please include all the emails returned from get_maintainer.pl including Greg's and Thomas's.

I already explained why is Cc list constructed the way it is. Despite formal rules, time is a scarce resource, so let's not draw people's attention too early. Anyway, if you think there's something that needs someone's special attention, you can still extend Cc list yourself.

Before sending v2, maintainer's review or ack is needed. I already collected Thomas' ack for a move, but have not read a single world from you. Do you plan to do some actual review, so we can take next step or are you willing to wait for a v2 which will add only octeon_get_io_clock_rate stub for non Octeon builds (having clk api would be nice here, but that's different story)?

Also, perhaps it would be reasonable to squash patches 8 and 9. Which tree to you want it to be rebased against? Currently I'm at linux.git master and changes were retested here.

I'd prefer to send fewer version if possible, so actual comments on code would certainly help.

[snip]

Quoted 8 lines
Quoted 5 lines

Also coleagues of mine meanwhile found that PLL indeed ocassionally fails to lock, so workaround attached to cover letter is really needed. Naturally it cannot sneak in as it is, so unless you have better idea I'll just port it to recent driver state and we can start discussion from there in a separate thread.

If this causes a regression, then please fix it before sending it in. If it's a new found issue, you can create a fix patch at a later point.

There are few problems with Octeon's DWC3 implementation and none of them can be really solved without documentation. Here I come in trouble as Sysnopsys tech support pointed to SoC manufacturer which no longer exists. Cavium was bought by Marvell, dumped almost all the staff and those still providing technical support were unable to find any revelant documentation in past few months.

So unless that changes I can send only hackish patches which kinda overcome issues found, but I do not think it is viable solution. That's something I'll do at later point as you suggested and we'll see if an acceptable solution pops up.

Thank you,
ladis

maintainer
#26 lore raw

Hi,

Quoted 10 lines · Ladislav Michl

Hi,

Quoted 1 line · Thinh Nguyen

Hi,

[snip]

Quoted 2 lines

For each patch, please include all the emails returned from get_maintainer.pl including Greg's and Thomas's.

I already explained why is Cc list constructed the way it is. Despite formal rules, time is a scarce resource, so let's not draw people's

Now that doesn't sound fair that you think my resource is more expendable. :)

Quoted 3 lines

attention too early. Anyway, if you think there's something that needs someone's special attention, you can still extend Cc list yourself.

I'm just asking you to follow the submitting patch process and add the appropriate maintainers on any code that they maintain. You just ignored the get_maintainer.pl output and only added what you feel needed.

Quoted 5 lines

Before sending v2, maintainer's review or ack is needed. I already collected Thomas' ack for a move, but have not read a single world from you. Do you plan to do some actual review, so we can take next step or are you willing to wait for a v2 which will add only

Please be patient, I'll get to your changes when I can, and I'm not the only reviewer here. Your series isn't exactly small or easy to understand for me to review quickly.

Quoted 9 lines

octeon_get_io_clock_rate stub for non Octeon builds (having clk api would be nice here, but that's different story)?

Also, perhaps it would be reasonable to squash patches 8 and 9. Which tree to you want it to be rebased against? Currently I'm at linux.git master and changes were retested here.

I'd prefer to send fewer version if possible, so actual comments on code would certainly help.

Well, you can help me and other reviewers by submitting v2 with your rebased changes. I think it's appropriate here given that this series touches on 2 separate subsystems.

Quoted 11 lines

[snip]

Quoted 8 lines
Quoted 5 lines

Also coleagues of mine meanwhile found that PLL indeed ocassionally fails to lock, so workaround attached to cover letter is really needed. Naturally it cannot sneak in as it is, so unless you have better idea I'll just port it to recent driver state and we can start discussion from there in a separate thread.

If this causes a regression, then please fix it before sending it in. If it's a new found issue, you can create a fix patch at a later point.

Quoted 6 lines

There are few problems with Octeon's DWC3 implementation and none of them can be really solved without documentation. Here I come in trouble as Sysnopsys tech support pointed to SoC manufacturer which no longer exists. Cavium was bought by Marvell, dumped almost all the staff and those still providing technical support were unable to find any revelant documentation in past few months.

This can be problematic...

Quoted 6 lines

So unless that changes I can send only hackish patches which kinda overcome issues found, but I do not think it is viable solution. That's something I'll do at later point as you suggested and we'll see if an acceptable solution pops up.

If it's hackish because of the lacking of documentation, then just note how you came to the solution in your patch is totally fine. There's nothing else we can do in that case.

I notice that there are many magic numbers in your changes. I hope you have documentations for them. I may need to make some assumption/skip reviewing some areas because of that.

Thanks,
Thinh